Edvinas
Navickas†
a,
Tobias M.
Huber†
a,
Yan
Chen
b,
Walid
Hetaba
c,
Gerald
Holzlechner
a,
Ghislain
Rupp
a,
Michael
Stöger-Pollach
c,
Gernot
Friedbacher
a,
Herbert
Hutter
a,
Bilge
Yildiz
b and
Jürgen
Fleig
*a
aInstitute of Chemical Technologies and Analytics, Vienna University of Technology, Getreidemarkt 9, Vienna A-1060, Austria. E-mail: J.Fleig@tuwien.ac.at
bDepartment of Nuclear Science and Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, 24-107, Cambridge, MA 02139, USA
cUniversity Service Centre for Transmission Electron Microscopy, Vienna University of Technology, Wiedner Hauptstr. 8-10, Vienna A-1040, Austria
First published on 9th January 2015
In this study, the contribution of grain boundaries to the oxygen reduction and diffusion kinetics of La0.8Sr0.2MnO3 (LSM) thin films is investigated. Polycrystalline LSM thin films with columnar grains of different grain sizes as well as epitaxial thin films were prepared by pulsed laser deposition. 18O tracer exchange experiments were performed at temperatures from 570 °C to 810 °C and subsequently analyzed by secondary ion mass spectrometry (SIMS). The isotope concentration depth profiles of polycrystalline films clearly indicate contributions from diffusion and surface exchange in grains as well as in grain boundaries. Measured depth profiles were analyzed by finite element modeling and revealed the diffusion coefficients D and oxygen exchange coefficients k of both the grain bulk and grain boundaries. Values obtained for grain boundaries (Dgb and kgb) are almost three orders of magnitude higher than those of the grains (Dg and kg). Hence, grain boundaries may not only facilitate fast oxygen diffusion but also fast oxygen exchange kinetics. Variation of the A-site stoichiometry ((La0.8Sr0.2)0.95MnO3) did not lead to large changes of the kinetic parameters. Properties found for epitaxial layers without grain boundaries (Db and kb) are close to those of the grains in polycrystalline layers.
Moreover, there is an open question regarding the bulk path in LSM: how much do grain boundaries contribute to the overall oxygen reduction rate? Some studies dealt with the effect of crystallographic orientation21 or substrate22 on ionic conductivity in LSM thin films. Different electronic properties of grain boundaries compared with grain surfaces were found by scanning tunneling microscopy (STM) studies on LSM layers.23,24 This also suggests a different catalytical activity for ORR at the grain boundaries, but quantitative data are not available yet. Impedance spectroscopic measurements on thin film microelectrodes can reveal information on the relative importance of bulk path and surface path11 and can even distinguish whether the bulk path for ORR is limited by the surface reactions or by transport through the bulk.14 However, information on grain boundary effects can hardly be resolved unambiguously by this technique.
18O exchange experiments followed by secondary ion mass spectrometry (SIMS) analysis – also called isotope exchange depth profiling (IEDP) – is a powerful technique to probe oxygen surface exchange and oxygen diffusion kinetics and/or to visualize the active ORR sites.25–34 There are also some IEDP studies on sintered LSM31,32,34 where it was found that grain boundary diffusion does have an influence on the overall oxygen transport in polycrystalline LSM samples. Quantitative information on the oxygen exchange properties of the grain boundaries was not deduced from those measurements. Additional information on oxygen diffusion in grains and grain boundaries of LSM, particularly in LSM thin films with active ORR bulk path, are therefore vital for understanding their role, especially in nanoscaled thin-film structures.
In this study, we report results of IEDP experiments revealing the strong acceleration effect of grain boundaries for oxygen surface exchange and oxygen diffusion in LSM thin films prepared by pulsed laser deposition. The 18O isotope exchange experiments were performed between 560 and 810 °C on polycrystalline LSM layers with different grain widths and on epitaxial layers. The isotope depth profiles were analyzed by numerical finite element modeling. In order to investigate the relevance of A-site non-stoichiometry, nominally stoichiometric (La0.8Sr0.2MnO3) and A-site deficient LSM ((La0.8Sr0.2)0.95MnO3) thin films were investigated. These measurements allowed for the first time the determination not only of the grain boundary diffusion coefficient but also of the grain boundary surface exchange coefficient kgb of LSM thin films.
All polycrystalline LSM layers were deposited on YSZ (100) (ZrO2 with 9.5 mol% Y2O3, CrysTec GmbH, Germany) single crystal substrates. The deposition was performed in oxygen atmosphere at 4 × 10−2 mbar oxygen pressure using a KrF excimer laser with a wavelength of 248 nm and a pulse frequency of 10 Hz. The laser beam energy was set to 400 mJ per pulse and a target-substrate distance of 6 cm was used. Different film growth temperatures ranging from 600 °C to ≥810 °C were employed to vary the grain width of the thin films with columnar grains. The substrate temperature was measured by an IR pyrometer (Heitronics KT-19.99, Germany).
Epitaxial LSM95 and LSM100 layers were deposited on SrTiO3 (STO) (100) (CrysTec GmbH, Germany) single crystals at 650 °C under 1.3 × 10−2 mbar oxygen pressure. The films were cooled down to room temperature in 2.7 mbar oxygen pressure. The other PLD parameters were the same as for the polycrystalline LSM layer preparation.
The thickness of the polycrystalline LSM layers was controlled by deposition time and later determined by scanning electron microscopy (SEM) (FEI Quanta 200 FEG, Netherlands) and transmission electron microscopy (TEM) (FEI TECNAI-20) from cross-section images. Values ranging from 170 nm to 212 nm were found. The grain width of LSM layers was estimated by two techniques: atomic force microscopy (AFM) (NanoScope V, Bruker Nano) and TEM. In the first case, the surface morphology of LSM95 and LSM100 thin films was characterized using AFM tapping mode. The resulting AFM micrographs were then analyzed by the linear intercept method. In TEM measurements the average column widths were determined from the available images.
The resulting 18O depth profiles were subsequently investigated by time-of-flight secondary ion mass spectrometry (ToF-SIMS) (ION-TOF GmbH, Germany ToF-SIMS 5). For these SIMS measurements the collimated burst alignment (CBA) mode with Bi3++ primary ions (25 keV) was used, which allows accurate determination of 18O fractions in a broad intensity range.35,36 Negative secondary ions were analyzed in areas of 30 × 30 μm2, using a raster of 256 × 256 measurement points. For the sputtering of material, 1 keV Cs+ ions were applied with a sputter crater of 300 × 300 μm2 and sputtering ion current of 70 nA. The charging of surfaces was compensated by an electron flood gun. The depth profiles of the isotope fraction (f(18O)) were obtained by normalizing integrated intensities I of 18O and 16O according to
(1) |
Fig. 1 θ–2θ scans of polycrystalline LSM95 and LSM100 thin films deposited at 695 °C on YSZ substrates. |
Owing to the similar lattice parameters of LSM and STO, preparation of LSM100 and LSM95 on STO substrates leads to epitaxial layer growth.21,37 The microstructure of the deposited layers was checked by 2θ–ω scans (Fig. 2a) and by reciprocal space mapping (Fig. 2b). The X-ray diffractograms indicate that both LSM layers are with (001) out-of-plane orientation and epitaxial. The small peaks in 2θ–ω scans around the (001) and (002) peaks (Fig. 2a) are thickness fringes, which come from constructive interference of X-rays reflected from the film and the substrate interface. The presence of such thickness fringes is an indication of the high quality of the film, with a smooth surface and film-substrate interface. As one can see from reciprocal space maps (RSM) (Fig. 2b), the peaks of LSM100 and LSM95 are located right beneath those of the STO substrate, which indicates that the in-plane lattice parameter of the LSM films epitaxially match to that of the STO substrate (3.905 Å). Keeping in mind that the LSM bulk lattice parameter (3.88 Å) (ref. 38) is smaller than for STO, this should lead to a tensile strain, which was also observed in another study.38
Layers formed at ≥810 °C have less pronounced microstructural features and in this case a grain width was estimated from TEM cross-section bright-field images. The TEM measurements were performed on LSM100 layers and examples are shown in Fig. 3. The measurements confirmed the columnar grain growth but also revealed some variation in grain width within a single layer. Keeping the limited statistics in mind, results found from TEM for LSM100 deposited at 600 °C and 700 °C are in a good agreement with AFM results, i.e. again ca. 30 and ca. 40 nm, respectively. The LSM100 layer deposited at ≥810 °C showed slightly larger grains in TEM and lead to an estimated value of 45 nm. For the LSM95 layer deposited at 810 °C the surface roughness is increased, but the films are still dense. Despite absence of TEM data for the LSM95 layer deposited at ≥810 °C we also expect a slight increase in grain size compared to the 700 °C layer. The following grain widths were therefore used in the quantitative analysis of the tracer profiles: 30, 40 and 45 nm for LSM100 layers and 60, 70 and 75 nm for LSM95 layers, respectively.
The grain size, manipulated by the film deposition temperature, also clearly affects the profiles. As an example, Fig. 4c and d displays the results for LSM100 layers prepared at 600 °C (d = 30 nm) and 860 °C (d = 45 nm) after oxygen exchange at 580 °C. The near-surface parts of the profiles are very similar while a larger grain size causes less 18O concentration in some depth. Profiles measured on epitaxial layers deviate strongly from those of polycrystalline layers. As shown in Fig. 4c, d and 5, the very sharp decrease is not followed by a long diffusion tail. At low temperature, natural abundance is almost reached at the LSM/substrate interface, despite very thin layers (28–29 nm). At high temperatures, however, very high tracer fraction levels penetrate into the films and the profiles become rather flat, despite short diffusion times. Pronounced diffusion in STO is not found.
Fig. 5 The 18O isotope exchange depth profiles of epitaxial LSM100 thin films obtained with different isotope exchange parameters. |
The absence of the shallow diffusion tail in epitaxial films and the dependence of this part of the diffusion profile on the grain size already indicate an effect of grain boundaries in the overall diffusion kinetics. Profiles strongly resemble those generally found in the type B regime of Harrison`s grain boundary diffusion classification.34,39 Similar isotope exchange depth profiles were also obtained in SrZrO3 pellets, where a fast grain boundary diffusion path was concluded.40 However, a straightforward analysis of the grain boundary diffusion coefficient according to the model by Whipple–Le Claire41 (analyzing lnf(18O) − x6/5 slopes) is not possible here: first, we do not have semi-infinite conditions in a single phase but a thin film and fast diffusion in the substrate. Second, we are probably not within the validity range of the model mentioned above, with the total amounts of tracer ions in grain boundaries not being negligible, and diffusion lengths in the grains partly being too large compared to the grain size.
Therefore, a numerical analysis of the measured 18O depth profiles by means of the finite element method (FEM) was performed (COMSOL). The 3D geometry of the model included three domains (Fig. 6), representing diffusion in a grain, along the grain boundary and in the YSZ substrate. The width of the grain boundary (δ) was set to 2 nm and the width of a grain (d) was varied in accordance with the approximative values found in the experiments (Section 3.2). For the sake of simplicity, a cylindrically shaped grain was considered, see r.h.s. in Fig. 6. The numerical tool solves Fick's second law of diffusion in all domains, with diffusion coefficients in LSM grains (Db), LSM grain boundaries (Dgb) and YSZ (DYSZ). The boundary conditions of the FEM model reflect that gas exchange is possible at the surface of grains (exchange coefficient kb) and grain boundaries (kgb). Diffusion thus takes place in the grain, along the boundary, from the grain boundary to the grain and from the LSM layer to the YSZ substrate. Since the profiles were too shallow in YSZ to deduce a diffusion coefficient, DYSZ used in the simulation was calculated from measured conductivity (σion) values.
Fig. 6 The sketch of the geometry used in the finite element model to analyze the measured isotope depth profiles. |
Then FEM calculations with varying D and k values of grain and grain boundary were performed until the best match was found between the calculated isotope profiles and the experimental ones, across the entire set of experimental data. As an additional parameter in the FEM model, the initial value of f(18O) was set to the natural abundance level of 0.00205 for 18O (given by the National Institute of Standards and Technology (NIST)) and the 18O concentration in the gas was assumed to be 97.1% (in accordance with the isotope supplier). The FEM calculations on epitaxial layers were done in a similar manner, though only with a single LSM domain, i.e. without grain boundary and without any diffusion in the STO substrate.
First, the concentration depth profile of a polycrystalline film was approximated by a single diffusion process (Db = Dgb and kb = kgb) (green line in Fig. 7a, b). However, in this manner only the steep part of the profile could be described. In a second step, different grain and grain boundary diffusion coefficients were used for the fit (Db ≠ Dgb and kb = kgb) (blue line in Fig. 7a, b). Again the measured profile could not be reproduced, since the very high 18O concentration in some depth cannot be explained by fast diffusion in the grain boundaries only. This suggests that also a modified surface exchange coefficient for the grain boundaries has to be involved in the model. Finally, full consideration of all fit parameters with different diffusion and different surface exchange coefficients (Db ≠ Dgb and kb ≠ kgb) leads to a successful fitting (black lines in Fig. 7). To the best of our knowledge this is the first quantification of the grain boundary surface exchange coefficient of a mixed conducting oxide.
The same type of FEM analysis of the experimental data was performed for all the polycrystalline samples, each with the corresponding grain size. Fig. 7c and d displays measurement data and fit results for two different grain sizes in linear and logarithmic plots, together with that of the epitaxial layer (see also below). The steep isotope depth profile part corresponding to the grain bulk diffusion is almost the same in all cases while deeper parts strongly depend on the layer microstructure.
The obtained diffusion and surface exchange values for the grain and grain boundaries are plotted in Arrhenius diagrams in Fig. 8. As one can see, both the diffusion coefficients and the surface exchange coefficients of the grain boundaries are orders of magnitude higher than those of the grain. The enhancement ratios are in the range of 200 to 1000 for Dgb/Db and 200 to 800 for kgb/kb. Moreover, the diffusion and surface exchange coefficients do not show a significant dependence on the grain size, all data are more or less on the same lines. This suggests that grain boundary widths and grain boundary structure/chemistry of all three layers do not drastically depend on the deposition temperature in the explored range. Moreover, this consistency of the data supports the validity of our proposed model with fast grain boundaries in LSM thin films. The strongly accelerated oxygen exchange and transport kinetics of grain boundaries might become an important contributor to the overall kinetics also in porous LSM electrodes with sub-μm grains, at least in regimes with relevant bulk path, e.g. upon significant cathodic polarization.
Still, k and D values slightly spread and the activation energies Ea obtained from Arrhenius plots have to be taken with some care. Fit lines for 30 and 45 nm grains are included in Fig. 8. The activation energies for both grain sizes were averaged and the resulting values for grain bulk (Db) and grain boundary diffusion (Dgb) as well as for both surface exchange coefficients (kb and kgb) are listed in Table 1. A more detailed discussion of activation energies is given below, together with values found for LSM95 and epitaxial layers.
E D b [eV] | E D gb [eV] | E k b [eV] | E k gb [eV] | |
---|---|---|---|---|
LSM100, polycrystalline | 1.70 | 2.22 | 1.53 | 1.88 |
LSM95, polycrystalline | 2.04 | 1.78 | 1.57 | 1.70 |
LSM100, epitaxial | 2.55 | 2.04 | ||
LSM95, epitaxial | 2.63 | 1.97 |
In the case of epitaxial layers and higher exchange temperatures, only the first part of the depth profile can be fitted with a single diffusion process (Fig. 9a). Obviously, again a second diffusion mechanism is present, even though these additional effects are much less pronounced than for polycrystalline layers. Owing to the supposed absence of grain boundaries in the epitaxial layers, one might attribute the shallow part of the profile to fast diffusion along other extended defects. Extended defects which may contribute to the diffusion of oxygen were also observed in other studies.22,37 However, in our FEM calculations only the bulk diffusion part was considered, since the density of such supposed additional defects in the LSM epitaxial layers was not known. The Db and kb values obtained from the epitaxial layer and averaged values of the polycrystalline LSM100 are plotted in an Arrhenius diagram (Fig. 9b). The similarity of the values supports our model assumption that the near-surface part of the profiles in the polycrystalline layers reflects the bulk diffusion behaviour. Slightly higher Db and kb values are found in the epitaxial layers, but statistical relevance of the differences is not proven yet. It might be related to tensile strain, which may change the kinetics,42 similar to results found for Sr-doped LaCoO3−δ.33
Comparing all activation energies in Table 1 one can conclude that activation energies of k values are smaller than those of D. Differences are of the order of 0.2 to 0.4 eV. A statement on any difference between activation energies of grains and grain boundaries of polycrystalline layers, however, is not possible yet, due to the scatter present in the fitted data. Approximately, we get 2 eV for D values and 1.7 eV for k values. Even the fact that activation energies of epitaxial layers are somewhat higher should not be over-interpreted. All these tracer measurements in thin films and their analysis are far from trivial and include some uncertainties such as very sharp profiles close to the surface at low temperatures, overlap of grain and grain boundary parts, i.e. almost linear profiles, at high temperatures (see Fig. 4a), and inaccuracies in determining the grain widths.
In order to check the effect of the grain size and the grain boundary width on the fitting results of an experimentally obtained depth profile of LSM95, the measured data were analyzed with different grain (d = 90, 75 and 60 nm) and grain boundary (δ = 0.5, 1 and 2 nm) widths in the FEM model. The resulting values are listed in Table 2. The variation of the grain width strongly affects the surface exchange coefficients of the grain boundary (kgb), while grain values (Db, kb) as well as Dgb are much less sensitive. This may explain the strong scatter of the surface exchange values in the LSM95 case, particularly since our model neglects any grain size distribution. The effect of the grain boundary width on kgb and Dgb is not surprising and in the range expected when assuming that δ·kgb and δ·Dgb are the decisive factors for the amount of incorporated tracer ions.
D b [cm2 s−1] | k b [cm s−1] | D gb [cm2 s−1] | k gb [cm s−1] | |
---|---|---|---|---|
Grain width d [nm] | ||||
90 | 2.1 × 10−16 | 3.3 × 10−10 | 3.9 × 10−13 | 1.8 × 10−8 |
75 | 2.2 × 10−16 | 3.4 × 10−10 | 3.9 × 10−13 | 1.5 × 10−8 |
60 | 2.2 × 10−16 | 3.6 × 10−10 | 3.9 × 10−13 | 1.0 × 10−8 |
Grain boundary width δ [nm] | ||||
0.5 | 2.2 × 10−16 | 3.4 × 10−10 | 8.0 × 10−13 | 5.0 × 10−8 |
1 | 2.2 × 10−16 | 3.4 × 10−10 | 4.0 × 10−13 | 2.4 × 10−8 |
2 | 2.2 × 10−16 | 3.4 × 10−10 | 3.9 × 10−13 | 1.5 × 10−8 |
Finally, the activation energies might be compared with those of a previous tracer diffusion study on polycrystalline LSM pellets [34]. There, activation energies for oxygen diffusion in bulk and grain boundaries in the temperature range of 700 °C to 1000 °C were 3.2 eV and 2.8 eV, respectively. In our study, both activation energies are lower. Regarding Dgb this might be due to different types of grain boundaries in thin films and polycrystalline pellets, caused by the columnar growth in thin films and the very different temperature treatments (deposition temperature much below typical sintering temperatures of pellets). This very different temperature treatment might also cause the severe differences between the absolute k values in ref. 32 and our kb, ours are two to three orders of magnitude larger. Moreover, in the corresponding analysis of polycrystalline LSM pellets, only fast grain boundary diffusion was included, while fast surface oxygen exchange of grain boundaries was not considered, which further hampers comparison of deduced grain boundary properties. Absolute values of Db, however, are very close in the temperature regime used in both studies (e.g. at ca. 800 °C). Hence, one could speculate that at about 700–800 °C a change of the activation energy of Db takes place and causes the different Ea values found in our study (at lower temperatures) and in ref. 34 (at high temperatures). However, also the very limited temperature range and fit inaccuracies at both ends of our temperature scale might play a role. A comparison of kgb to literature values is not possible since, to the best of our knowledge, this factor has never been measured before.
Footnote |
† E. Navickas and T. M. Huber contributed equally to this work. |
This journal is © the Owner Societies 2015 |