Apparent Oxygen Uphill Diffusion in La0.8Sr0.2MnO3 Thin Films upon Cathodic Polarization

Abstract The impact of cathodic bias on oxygen transport in La0.8Sr0.2MnO3 (LSM) thin films was investigated. Columnar‐grown LSM thin films with different microstructures were deposited by pulsed laser deposition. 18O tracer experiments were performed on thin film microelectrodes with an applied cathodic bias of −300 or −450 mV, and the microelectrodes were subsequently analyzed by time‐of‐flight secondary ion mass spectrometry. The 18O concentration in the cathodically polarized LSM microelectrodes was strongly increased relative to that in the thermally annealed film (without bias). Most remarkable, however, was the appearance of a pronounced 18O fraction maximum in the center of the films. This strongly depended on the applied bias and on the microstructure of the LSM thin layers. The unusual shape of the 18O depth profiles was caused by a combination of Wagner–Hebb‐type stoichiometry polarization of the LSM bulk, fast grain boundary transport and voltage‐induced modification of the oxygen incorporation kinetics,


Introduction
La 0.8 Sr 0.2 MnO 3 (LSM) and similarp erovskite-type materials are widely investigated for solid oxide fuel cell cathode applications. [1] Owing to its low ionic conductivity,L SM is often considered to be at hree-phase boundary active material, [1c] and the bulk path, that is, oxygen reduction with ion transport throughL SM, is relevant only in thin films. [2] However,t he oxygen reduction reaction at the surfaceo fL SM and oxygen diffusioni nL SM can be varied by an applied cathodic bias. [2a, 3] The bias dependence of oxygen diffusion is due to stoichiometry changes in LSM upon polarization caused by am odified chemicalp otential of oxygen.H igher oxygen vacancy concentrationsr esult and can improvet he electrochemical performance of LSM electrodes. Therefore, the bulk path of oxygen reduction may be highly important not only in thin films but also in polarizedp orous LSM cathodes. [4] Simulations of the relevance of the bulk path in polarizedL SM electrodes are presented in Ref. [4b].A na pplied cathodic bias further affects the oxygen incorporation rate at the surface, thoughd etails of these changes ando ft he oxygen incorporation mechanism in LSM are not yet well understood. Additional experiments on LSM electrodes under operating conditions are therefore neededt oo btain ac learp ictureo ft he kinetics of oxygen reduction occurring through the bulk path. Thin films are particularly useful in this respect, due to the enhanced relevance of the bulk path, their simple geometry,a nd the accessibility of the surface to surface analytical tools. 18 Ot racer diffusion and subsequent secondary ion mass spectrometry (SIMS) analysisi sapowerful technique that allowso xygen exchange and oxygen ion diffusiont ob e probeda nd/ort ov isualize the active oxygen reduction sites. [5] In previous studies,b ias-induced 18 Oi ncorporation experiments on LSM were successfully employed to qualitatively show the relevance of the bulk path for oxygen reduction. [5b, c] Recent contributions on thermalo xygen tracer incorporation into LSM thin films revealed further detailso nt he mechanism of oxygen surfacee xchange and diffusion. [5f, 6] It wass hown that the grain boundaries play am ajor role and have diffusivities and surfacee xchange coefficients that are orders of magnitude higher than those of the grain bulk. Quantitative information on how oxygen incorporation and diffusion in LSM thin films is affected by ac athodic biasiss till missing.
The goal of this work is to reveal the effect of ac athodic bias on oxygen incorporation and diffusion in LSM thin films by tracer experiments and subsequent depth profiling. To show the role of the thin film microstructure, layers with different grain widths were investigated. Bias-driven isotopei ncorporation was performed on LSM microelectrodes with an applied cathodic biaso fÀ300/À450 mV in the temperature range of 500 to 700 8C. The isotope depth profiles were measured by SIMS, and additional finite element modeling (FEM) revealed the highr elevance of fast grain boundary transport but also the relevance of defect concentration gradients for oxygen reduction on LSM thin films under cathodic bias.
The impacto fc athodic bias on oxygen transport in La 0.8 Sr 0.2 MnO 3 (LSM) thin films wasi nvestigated. Columnargrown LSMt hin films with different microstructures were depositedb yp ulsed laser deposition. 18 Ot racer experiments were performed on thin film microelectrodes with an applied cathodic bias of À300 or À450 mV,a nd the microelectrodes were subsequently analyzed by time-of-flight secondary ion mass spectrometry.T he 18 Oc oncentration in the cathodically polarizedL SM microelectrodes was strongly increased relative to that in the thermally annealed film (withoutb ias). Most remarkable, however,w as the appearance of ap ronounced 18 O fraction maximum in the center of the films. This strongly depended on the appliedb ias and on the microstructure of the LSM thin layers. The unusual shape of the 18 Od epth profiles was causedb yacombination of Wagner-Hebb-type stoichiometry polarization of the LSM bulk, fast grainb oundary transport and voltage-induced modification of the oxygen incorporation kinetics,

Results
Twos ets of LSM samples were prepared at differentd eposition temperatures (600 and 900 8C), which led to different grain widths of the columnar grown films. As shown in Ref.
[5f] by using atomic force microscopy (AFM) and transmission electron microscopy (TEM), LSM layers deposited at 600 8Ch ave grain diameters of about 30 nm and those deposited at 830 8C consist of grains that are roughly two times larger.F igure 1a,b displays the surface topography of the LSM layers measured by AFM. The LSM layer deposited at 600 8C( LSM 600 )s hows as urfacet opography with narrow and well-definedg rain width ( % 30 nm, Figure 1a). However,t he LSM layer deposited at 900 8C( LSM 900 )h as coarse surface features and much higher average surface roughness (Figure 1b). This hindered determination of the exact grain width,b ut we concluded that this sample hadavery different microstructure with grains larger than those found in LSM 600 .
Incorporation of the 18 Oi sotope was performed on rectangular LSM microelectrodes with an applied cathodic bias U of À300 or À450 mV.I nc ontrast to anodic voltages of similar magnitude, this cathodic polarization does not lead to visible morphological changes in the electrodes (no partial detachment). [2a] Am icroelectrodew as electrically contacted in as ymmetrically heated isotope exchange chamber (Figure 1c,d), which helped to avoid inhomogeneous temperature distribution within the microelectrode. [5a, 7] Immediately after 18 O 2 exposure,t he contacted microelectrode was polarized and the dc current wasm onitored during the entire experiment. Each sample had numerous microelectrodes, and thus, unpolarized LSM wasa lso exposed to 18 O 2 ,w hich allowed comparison with the thermal diffusion sample. An overview of all polarized electrodesi sg iven in Ta ble 1w ith the deposition temperature (T dep ), tracer exposure temperature (T ex ), bias voltage U,c athodic overpotential h,c urrent density found after 10 min of polarization,a nd tracer fraction at the LSM/yttria-stabilized zirconia (YSZ) interface.
Ta ble 1i ndicates that, as expected, the measured current increasesw ith measurement temperature and cathodic bias. Moreover,i tw as found that the LSM deposition temperature (T dep )a lso plays as ignificant role for the bias-driven current. Typical current versus time graphs for an applied cathodic bias (À300 mV) are shown in Figure 1e for both deposition temperatures.F irst, LSM electrodes with narrow grains (LSM 600 )e xhibit am uch higher current density for the same applied voltage and the same experimental conditions than an electrode with wider grains (LSM 900 ). Thus, the total electrode polarization resistance varies in accordance with the grain size. This can be understood from ap revious study on very similarL SM films with thermally driven 18 Od epth profiles. [5f] There, it was found Figure 1. AFM micrographs (1 mm1 mm) of LSM thin films prepared at a) 600 8Cand b) 900 8Cshowing different microstructures. c) Sketch of asymmetrically heated measurement/gas-exchange set up that includes ag astight quartz chamber placed in at ubefurnace, ac ontact arm with acontact needle, and as ample holder;aphotograph of ac ontactedm icroelectrode is shown on the right-hand side. d) Sketch of LSM microelectrodes on aY SZ substratewith electricalcontact. e) The typical current responsew ith an applied bias (À300 mV) obtained at 700 8Co nm icrostructurally different layers (El7 depositedat9 00 8Ca nd El9 deposited at 600 8C). www.chemelectrochem.org that the diffusioni nL SM grain boundaries wasa pproximately three orders of magnitude faster than that in LSM grains and also that surface oxygen exchange coefficients were much larger for grain boundaries;t he more grain boundaries the higher the effective oxygen exchange rate of the film. Therefore, the current is higher for films with small grains. Second, the different grain widths affect the shape of the current decay.I rrespective of the applied bias, the measured current response of all microelectrodes deposited at 600 8C shows an exponentiald ecay function (Figure 1e, c c). The current microelectrodes depositeda t9 00 8Ce xhibit additional time-dependent features (Figure 1e, c c): af ast current decrease is followed by an increase and ap lateau.T his indicates that at least two processes with different time constantst ake place, possibly stoichiometry polarization of grains and grain boundaries. Am ore detailedi nterpretationi sb eyond the scope of this paper;h ere, we only concludet hat the microstructure strongly affects the electrochemical properties of LSM films.
At ypical tracer depth profile obtained on LSM electrodes withoutappliedbias consists of two parts:asteep near-surface decrease in the 18 Oc oncentrationt hat is followed by as hallow decay of isotope fraction (Figure 2a, *). In accordance with Ref.
[5f],t his indicates two paralleld iffusion processes:T he first profilep art is dominatedb yd iffusion in the bulk of the LSM grains. The following long tail up to the LSM/YSZ interface is caused by fast grain boundary diffusion and continuous tracer "leakage" into the bulk, compare type Bd iffusion in Harrison's classification of grain boundary diffusion. [8] The concentration at the LSM/YSZ interface is above the natural abundance of 18 O, and av ery shallow profile is visible in YSZ, which indicates very fast diffusion in YSZ. The tracer ions in YSZa re more relevant for proper data analysist han one might expect from their low level, as their total amount in the 500 mmt hick YSZ singlec rystal can become quite significant.
To investigate whether the uphill diffusion shape and the high absolutei sotope level werei ndeed caused by an applied cathodic bias, isotope incorporation experiments were performed at al ower temperature (T ex = 500 8C) on microstructurally identicalL SM layers by applying cathodic biases of À300 and À450 mV.T he resultingt racer profiles under different cathodic biases are shown in Figure 2b.H igher cathodic bias leads to substantially higher tracer fractionsi nL SM and an increase of the near surface tracer fraction from 7.5 to 14 %. The uphill profile is very pronouncedfor À450 mV with amaximum at ad epth of 85 nm butt ransforms into ak ind of plateau( up to 40 nm depth) for À300 mV.
The effect of the diffusion temperature is shown in Figure 2c for LSM electrodes exposed to 18 O 2 between 500 and 700 8C. The resultso btained at higher temperatures indicate av ery pronounced "uphill diffusion". Given that the depth of the peak value shifts towards the surface, from about 80 nm depth at 600 8Ct or oughly 60 nm depth at 700 8C, the slope of this uphill part is even larger at 700 8Ca tÀ300 mV.F urthermore, the first steep tracer decrease, whichf or thermalo xygen diffusion is attributedt ob ulk diffusion, is disguised at the higher temperature by the uphill profile.
The clearest indication of the mechanism behind the unusual diffusionp rofiles comes from isotope incorporation experiments performed on microstructurally different LSM 600 and LSM 900 layersw ith ac athodic bias of À300 mV at temperatures of 600 and 700 8C. Changing from aL SM 600 (Figure 2d,v iolet symbols) to aL SM 900 (Figure 2d orange symbols) microelectrode with much larger grains leads to ad rastic change in the depth profile. At 600 8C, the near surface feature is essentially the same for both LSM microelectrodes. The "uphilld iffusion" part with its maximumn ear the center of the LSMf ilm, however,i sm uch less pronouncedf or LSM 900 than it is for the electrode with the small grains (LSM 600 ). Also, the overall isotope level is much lower in LSM with al ower density of grain boundaries.T his strongly suggests ac rucial role of grain boundary diffusion.

Discussion
To explain the diffusion profiles, particularly their unusual shapes with apparent uphill diffusion, we have to discuss the defectc hemical effects occurring after applying ab ias voltage. First, we consider the changes in the bulk of am ixed conducting electrode upon polarization.T he overpotential h leads to as patially varying chemical potential of oxygen m O .T he two extreme cases are given by oxygen reduction by ab ulk path with rate-limiting surfacek inetics (e.g. found for Sr-doped LaCoO 3Àd electrodes) [9] and ab ulk path with rate-limiting ion transport. In the first case, polarization leads to as tep in m O at the surfacea nd ac onstant m O in the electrode. In the second case, the situation corresponds to Wagner-Hebbpolarization [10] with electron blockinga tt he electrode/YSZ interface. Hence, the chemicalp otential of oxygen varies within the electrode. From earlier tracer andi mpedancem easurements [11] we know that for LSM both surface exchange kinetics and oxygen bulk transport are relevant,a nd thus, ac hemicalp otential distribution with surfaces tep and bulk decay results (see Figure 3a). Assuming negligible changes in the electronic majorityc hargecarrier concentration and, thus, ar ather constanta nd polarization-independent chemical potentialo fe lectrons (m e ), we get from m O + m V + 2 m e = 0( m V = chemical potential of oxygen vacancies)[ Eq. (1)]: in which R and T denotethe gas constant and temperature, respectively.This meansthat upon bias, the oxygen vacancy concentration in LSM may drastically change. This can be quantified by considering the chemical permeability [Eq. (3)]: with ionic and electronic conductivities s ion , s eon ,t hat determine the vacancy flux density J V by [Eq. (4)]: [12] J in which D V and u V are the vacancy diffusion coefficient and mobility,r espectively,a nd s ion = 2F·c V ·u V (F = Faraday's constant).
In the steady state, J V is constant, and al inear vacancy concentration profile with ah igher vacancy concentration at the LSM/YSZ interface results( see Figure3a). Accordingly,t he bulk tracer diffusionc oefficient D b also varies in the LSM film, as it is proportional to the oxygen vacancy concentration by [Eq. (5)]: with correlation factor f c . This consideration of bulk defect chemistry upon polarization, together with fast oxide ion diffusion along grain boundaries,a lready qualitativelye xplains the observed profile shape. The local depth-dependentc hemical potentialm odifies the tracer bulk diffusion coefficient in LSM such that diffusion in the LSM grains is relativelys low close to the surface and becomes faster towards the LSM/YSZ interface. This affects the shape of the bulk diffusion profile (whichi sp articularlyr elevant near the surface) but is even more important for "leakage" of the tracer from the fast grain boundaries into the grain. The in-plane transport coefficient of ag rain strongly increasesw ith depth. This is indicated by red arrows in Figure 3b.H owever,t he driving force fort racer "leakage" from af ast grainb oundary into the grain strongly depends on the tracer fraction in the grain boundary.N ear the surface, the 18 O concentration in the grain boundaries is highest, and the blue arrows in Figure 3b indicate that the 18 Of raction decreases with depth.
Qualitatively, the product of the transport coefficient (represented by red in-plane arrows) and driving force (represented by blue across-plane arrows)d etermines the resulting tracer fraction at ac ertain depth.N ear the surface, as maller fraction is incorporated from the grain boundaries into the grains due to the low bulk diffusion coefficient. Thed iffusion coefficient in the grains is strongly enhanced at some depths, and thus, more 18 Oi si ncorporated from the grain boundaryi nto the grain. The tracer fraction in the grain boundary becomes low near the LSM/YSZ interface, and again,l ess tracer transfers from the grain boundary to the grain, despite the high bulk diffusion coefficient. This should lead to in-plane isotope diffusion profiles in grains as sketched in Figure 3b,a nd exact shapes are discussed in the finite elementm odeling part below.O ur SIMS measurements cannotr esolve the lateral( inplane)p rofiles within as ingle grain but integrate over many grains. In parallel to this grain boundary diffusion with leakage into the grain, across-plane bulk diffusion originating at the surfacetakesplace. Close to the surface this adds an additional tracer fraction with as harp decay due to slow bulk diffusion.
Accordingly,w ec an expect exactly the profile shape found in the experiments ( Figure 3c): as harp drop closet ot he surface and am aximum oxygen fraction at some depth due to very pronounced tracer "leakage". The chemical potentialvariation sketched in Figure 3a is expected to also vary the vacancy concentrationa nd, thus, the tracer diffusion coefficient in the grain boundary.H owever,t his should only modify the exact tracer concentrationp rofile along the grain boundary,t hat is, the decay function of the driving force fort racer leakage( blue arrows in Figure 3b), but does not alter the main considerations.
On the basis of these assumptions, profiles were also modeled by finite element calculations (COMSOLM ultiphysics) for ac ylindricallys haped grain. The model includes three domains representing diffusion in ag rain (D b ), along the grain boundary (D gb ), and in the YSZ substrate (D YSZ ). Diffusion coefficient values in YSZ (D YSZ )w ere taken from conductivity measurements.M oreover,t wo different oxygen surface exchange coefficientsf or grain (k b )a nd grain boundary (k gb )w ere considered in the model. As ketch of the model used for the calculations is shown Figure 4a.A si nitial or boundary condition, the natural abundance in the sample was set to 0.00205 (given by the NationalI nstituteo fS tandards and Technology) for t = 0a nd the 18 Of raction during the experiment was set to 97.1 %( as provided by tracer gas supplier).
First, modeling was performed to describe the measured thermald iffusion profiles; details are described in Ref.
[5f].I n this case, four individual parameters (D b ¼ 6 D gb and k gb ¼ 6 k b ,a ll withouta ny depth dependence) allow successful fit to experimental tracer profiles (Figure4c, * and green line). To minimize unknownp arameters, the mean grain width of such films was set to 30 nm for LSM 600 ,and the width of grain boundaries in LSM was fixed to 2nm. The fit parameters of at hermal isotope diffusionp rofile are listed in Table 2a nd reveal diffusion (D)a nd surfacee xchange coefficients (k)f or the grain boundaries (gb) that are about three orders of magnitude highert han those forthe grains (b).
Oxygen tracer motioni ne lectrochemically polarized oxides consists of two flux terms;o ne that describes standard tracer diffusion (i.e. counter diffusion of 18 Oa nd 16 Oi ons) and one that represents the unidirectional ionic current flux through the entire sample. Ag eneral discussion of this combination of fluxes is given in Ref. [12].T here, it is shown that in the steady ChemElectroChem 2015ChemElectroChem , 2,1487ChemElectroChem -1494 www.chemelectrochem.org state with ac urrentd ensity j in the across-plane( z)d irection, Equation (6) has to be solved.
The symbols c18 O and c total denote the 18 Ot racer and total oxide ion concentrations,r espectively.I no ur case, we have to consider as patially varyingt racer diffusionc oefficient D and different current densities in grain boundaries and grains. In Ref. [12] it is also discussed that the relevance of the second flux term in Equation (6) (proportional to j)s cales with the 18 O tracer traction and is small relative to the diffusional term for 18 Of ractions below 10 %.
It is beyond the scope of this paper to analyze our measurement data quantitatively,a nd thus, we restrict our modeling to the first (standard diffusion) term in Equation (6) even though tracer fractionsi nt he 30 %r ange are found. Accordingly,f or voltage-driven tracer depth profiles the performed FEM calculations are similar to those without ac urrent. Only spatial variation of the diffusion coefficient due to stoichiometry polarization is introduced. In accordance with Equation (4) and J V = constant (steadys tate), the bulk grain tracer diffusion coefficient varies linearly according to [Eq. (7)]: in which h denotes the LSM film thickness (in this case À180 nm) and D represents the enhancement factor of the grain diffusion coefficient relative to the value at the surface [D b (z = 0)]. Only voltage-induced variation of the bulk diffusion coefficient D b was assumed for simplicity; the (larger) grain boundary diffusion coefficient D gb was still assumed to be depth independentinthese calculations. The simulation was performed in as teplike process. The calculated profile was first adjustedt ot he steep near-surface region of the exemplary measured data (LSM 600 , T ex = 600 8C, À300 mV) by changing D b and k b .A ccordingly,t his part of the profile was again attributed to slow oxygen incorporation into the grain and slow oxygen diffusion. The values of D gb and k gb were then chosen to reacht he measured 18 Oc oncentrationl evel at the LSM/YSZ interface. At ad epth of about 10 nm, at which the measured uphill diffusion starts, the grain boundary contribution becomes visible in the profile. This feature was finally adjusted by modifyingt he enhancement factor D as depicted in Figure 4c.
In this manner,p rofiles with apparent uphill diffusion can easily be reproduced, and the calculations confirm our qualitative interpretation. The in-plane isotope profiles obtained for D = 30 are shown in Figure 4b.N ear the grain surface, al ow value of D b limits the diffusion from the grain boundary to the grain;t herefore, "small" integrals of the isotope fraction are found for cross sections S0 and S1. At somed epth, D b becomes larger and more oxygen diffuses from the grain boundary into the grain (highest integrated amounts for S2 and S3). Approaching the interface,ahigh value of D b is found but al- Figure 4. a) Sketch of the cylindrical finite element model that consists of three domains: LSM grain (defined by D b and k b ), grainboundary (defined by D gb and k gb ), and YSZ substrate (definedb yD YSZ ). D b varies linearly from the surface to the LSM/YSZ interface, D gb was constant.b)In-planet racer fraction profiles in different depths (S0 to S5, indicatedina )calculatedf or D = 30. c) The experimental 18 Oi sotope depth profileso ft hermalo xygen diffusion at T ex = 600 8C(*)and bias-based transportatÀ300 mV (*). In the bias case, five solutions of the finite element( FEM) calculations are shown, for which D was varied from 0to5 0. Table 2. Parameters for nonpolarized and polarized (À300 mV) LSM microelectrodes( El1, T ex = 600 8C) used in the finite elementcalculations that are showninF igure 4.

Bias [mV]
D 08.0 10 À14 1.5 10 À8 4.0 10 À17 2.3 10 À11 -À300 2.3 10 À12 2.0 10 À7 1.6 10 À16 8.5 10 À11 0-50 ChemElectroChem 2015, 2,1487 -1494 www.chemelectrochem.org ready much less isotope is availablei nt he grain boundary (S4 and S5). Thus, in-plane profiles become rather flat and include less tracer ions. However, within the framework of the given model the exact shape and absolute height of the measured curve cannotb ef itted accurately.A lso, additional consideration of the depth dependence of D gb is not sufficient for an accurate fit, as the second term in Equation (6) is neglected. Moreover, the currentv ersus time measurements (Figure 1e)a lready indicated that steadys tate is not established for as ignificant part of the overall diffusion time. This is an unavoidable consequenceo ft he gas exchange process (see the Experimental Section). Hence, additional time dependencies and parameter modifications come into play:1 )Establishing the steady-state profile of m O takes somet ime (given by chemical diffusion and chemicals urfacee xchange coefficients). During this period,a ll parameters, that is, k b , k gb , D b (z), and D gb (z), are time dependent. 2) This time dependence mostp robably includes different timescales, as chemical diffusion is also expected to be faster along grain boundaries, and thus, stoichiometry polarization of the grains takes place not only from the surface( in the z direction) but also from the grain boundaries with temporal in-plane vacancy concentrationv ariation.
Implementing all these additional aspects into the model would be required to finally quantify the measured profiles and to deduce information on the exact bias dependence of the k values and the contributionofD. The latter also indicates how much of the driving force h is reflected by as urfaces tep of m O and by rm O in the grain. Thisd etailed analysis is beyond the scope of this paper and requires furthere xperimentation. However,w em ay still discusst he parameters found for the very qualitative "fit procedure" done so far ( Table 2). Allp arameters are enhanced upon bias voltage, that is, the k factors of the bulk andg rain boundaries are substantially larger and also the D(z = 0) value in the grain is highert han that for thermal diffusion.H ence, ac ertain step of m O at the surface is present, which indicates combineds urface/transport rate limitation. According to the preliminary fit, concentration enhancement factors of several tens are most realistic for À300 mV.T his would also be in agreement with the upperl imit of the D value (D max ) realized for a m O curve withoutsurface step, that is [Eq. (8) and thus D max ¼ e 2Fh=RT ¼ 2901 for À300 mV at 600 8C.

Conclusions
Defect chemical processes and ion transport in polarized LSM microelectrodes werei nvestigated by means of voltage-driven 18 Ot racer gas incorporation and subsequent SIMS analysis. The measured dc current was enhancedu pon reducingt he LSM grain size andt hereby upon increasing the contributiono f grain boundaries to oxygen reduction. Oxygeni sotope depth profiles of voltage-driven 18 Oi ncorporation were characterized by very uncommon uphill-like diffusion with a 18 Ot racer maxi-mum in the center of the LSM film. This effect was causedb y the interplay of fast oxide ion diffusion along grain boundaries and stoichiometry polarization of LSM upon application of av oltage, which led to av acancy concentrationg radient in the LSM grains. This was particularly pronouncedf or LSM films with small grains. Numerical finite elements imulations confirmed that oxygen transport in two parallel paths, that is, by grains and grain boundaries, can lead to apparent uphill diffusion profiles. Preliminary quantitative analysisi ndicates that surfacei ncorporation kinetics andd iffusion both contribute to the rate limitation in polarized LSM microelectrodes and that both are accelerated by an applied cathodic bias in grains as well as in grain boundaries.

Experimental Section LSM Thin Film Deposition
Columnar LSM thin films were prepared by pulsed laser deposition (PLD). The PLD target was made from La 0.8 Sr 0.2 MnO 3 powder (Sigma-Aldrich, USA), which was isostatically pressed and sintered for 12 ha t1 200 8Ci nair.T hin LSM films were deposited on polished YSZ (100) single crystals (9.5 mol %Y 2 O 3 ,C rysTec GmbH, Germany) by using aK rF excimer laser (l = 248 nm, COMPex Pro 101 F, Lambda Physics, Germany). Laser beam energy was set to 400 mJ per pulse at 10 Hz pulse frequency.T he deposition was performed under an O 2 atmosphere (4 Pa) and with at arget-substrate distance of 6cm. To vary the microstructure of the LSM layer,t wo deposition temperatures (T dep )o f6 00 8C( LSM 600 )a nd 900 8C (LSM 900 )w ere used, which was monitored by ap yrometer (Heitronics KT-19.99, Germany). The film thickness was controlled by ak nown deposition rate. Squared 490 490 mm 2 or 390 390 mm 2 LSM microelectrodes were prepared from these films by UV photolithography and chemical etching in concentrated hydrochloric acid. Ap latinum counter electrode was brushed on the back side of the YSZ substrate. The grain widths and surface topography of the LSM thin films was checked by atomic force microscopy (AFM, Nanoscope V, Bruker Nano). Samples with microelectrodes were divided into smaller pieces, which thus allowed several samples with the same LSM film thickness and microstructure to be investigated.

Oxygen Tracer Incorporation upon Cathodic Polarization and Profile Analysis
AL SM microelectrode was contacted by aP t/Ir tip (see Figure 1), and the quartz tube with the sample was moved into at ube furnace. [7] After thermally equilibrating this gas-exchange set up in air at temperatures (T ex )o f5 00, 600, and 700 8C, the system was evacuated to ap ressure of roughly 1Pa. An 18 O 2 tracer gas atmosphere (200 hPa, 97.1 %, Campro Scientific, Germany) was then filled into the sample chamber and immediately ac athodic bias of À300 or À450 mV was applied to the contacted LSM microelectrode by means of aP OT/GAL 30V 2A test interface together with an Alpha-AH igh Resolution Dielectric Analyzer (both Novocontrol, Germany) in ad cm ode (software WINCHEM and WINDETAN ovocontrol, Germany). The dc current was monitored during the entire experiment. Given that each sample had numerous microelectrodes, unpolarized LSM was also exposed to 18 O 2 ,which allowed comparison with thermal diffusion. The 18 Of raction in the gas-exchange chamber was checked by am ass spectrometer (Pfeiffer GSD320 with QMG 220, Germany) and was in agreement with the 18 Of raction ChemElectroChem 2015ChemElectroChem , 2,1487ChemElectroChem -1494 www.chemelectrochem.org given by the supplier.T he cathodic overpotential h of the microelectrodes was determined by subtracting the voltage drop in the electrolyte from the applied bias voltage U.T he electrolyte resistance was measured by impedance spectroscopy;t he much smaller overpotential of the large counter electrode was neglected. All bias voltages and overpotential values are summarized in Ta ble 1; in the text and figures, only the total bias U is indicated. Isotope incorporation experiments lasted 10 min, and afterwards, the quartz tube was moved out of the tube furnace and the sample was quenched under an 18 O 2 atmosphere (cooling rate:1 00 8Cmin À1 ).
Ap re-annealing step of the thin films prior to the isotope experiments would be beneficial to chemically equilibrate LSM and thus to avoid chemical diffusion. [5e, 13] Also, establishment of ac urrent steady state prior to tracer exposure would be helpful if ab ias voltage is applied. However,i no ur experiments neither pre-equilibration nor steady state was possible, as the gas switch from ambient air to oxygen isotope gas required evacuation of the exchange chamber.T his step annihilates any chemical pre-equilibration or steady state. Gas exchange at room temperature was also not an alternative due to the very short tracer exposure times needed for thin films and the finite heat-up time. Hence, ac ontribution of chemical diffusion to the tracer experiment could not be avoided.
The resulting 18 Od epth profiles were subsequently investigated by time-of-flight secondary ion mass spectrometry (ToF-SIMS 5, ION-TOF GmbH, Germany). Measurements were done in the collimated burst alignment (CBA) mode with Bi 3 ++ primary ions (25 keV). This mode allowed accurate determination of 18 Of ractions over abroad intensity range. [14] Negative secondary ions were analyzed in an area of 45 45 mm 2 .F or the sputtering of material, 2keV Cs + ions were applied with as putter crater of 350 350 mm 2 and sputtering ion current of 120 nA. Surface charging was compensated with an electron flood gun. The tracer fraction f( 18 O) was obtained by normalizing integrated intensities I of 18 Oa nd 16 Oi nt he mass spectra according to Equation (9) The sputtering rate of LSM thin films was determined from the depth of as puttered crater that was measured by digital holographic microscopy (DHM, Lyncee Te c, Switzerland). The isotope depth profile measurements were performed on biased and on nonbiased microelectrodes, which thus allowed thermally and bias-driven oxygen tracer diffusiontobep robed.