Visualizing local fast ionic conduction pathways in nanocrystalline lanthanum manganite by isotope exchange-atom probe tomography of Chemistry 2022

The call for material systems with enhanced mass transport properties is central in the development of next-generation fuel cells, batteries and solid state energy devices in general. While two-dimensional doping by arti ﬁ cial heterostructuring or nanoscaling has shown great potential for overcoming kinetic limitations of ion di ﬀ usion, the length scale of interface e ﬀ ects requires the development of advanced tools for capturing and quantifying local phenomena in greater detail. In the present paper, an in-depth study of grain boundary oxygen conduction in Sr-doped lanthanum manganite ﬁ lms is presented by means of novel isotope-exchange atom probe tomography. Local pathways for fast masstransportaredirectlymappedbytwo-dimensionalreconstructions and line pro ﬁ les of the oxygen isotope concentration. Accurate ﬁ nite element modelling is employed to retrieve the local kinetic parameters, highlighting an enhancement of two orders of magnitude for both the di ﬀ usivity and surface exchange rate with respect to the bulk ( D * gb ¼ 2 : 1 (cid:2) 10 (cid:3) 14 cm 2 s (cid:3) 1 and k * gb ¼ 4 : 3 (cid:2) 10 (cid:3) 10 cms (cid:3) 1 ,respectively,for grain boundaries at 550 (cid:4) C). Co-acquired reconstruction of the cationic distribution reveals strong inhomogeneities (dopant de-mixing) across the grain boundaries and in the sub-surface region, leading to local Sr accumulation. The ﬁ ndings provide unequivocal quantitative assess-ment of fast grain boundary oxygen di ﬀ usion in lanthanum strontium manganite,givingfurtherinsightsintolocalstoichiometrydeviationsand promoting isotope exchange-atom probe tomography as a powerful tool for thestudy oflocalinterfacee ﬀ ects with high resolution. Di ﬀ erent models for the explanation of the phenomena are critically discussed on the basis of the experimental ﬁ ndings.


Introduction
The modication of the local chemistry via the introduction of an interface is a strategy with great potential for engineering the electrochemical properties of functional oxides beyond the state of the art (SoA). [1][2][3] In mixed ionic-electronic conductors (MIECs), which nd application e.g. as oxygen reduction electrocatalysts in solid oxide cells as well as in gas separation membranes, 4,5 solid solution is conventionally applied for increasing the electronic and oxygen vacancy concentration for oxygen exchange and diffusion. 6 Such an approach is however inherently limited e.g. by dopant solubility, by the reduction of ionic mobility caused by lattice disorder and defect association and, in some cases, by the high enthalpy of oxygen vacancy formation. [7][8][9] Due to the resulting limited ionic conductivity, the technological application of MIEC materials (e.g. as fuel cell cathodes) is based on MIEC/oxygen conductive uorites (doped ceria or zirconia) composites. [10][11][12] Enhancing the bulk conductivity of MIECs is therefore of primary importance for activating the bulk reaction and oxygen transport path and providing a potentially superior alternative to triple phase boundarybased SoA composites. Recent works have demonstrated a substantial enhancement of surface exchange rate and oxygen diffusivity, with respect to the bulk, for example at grain boundaries (GBs) of La 0.8 Sr 0.2 MnO 3 (LSM), a model interfacedominated MIEC. 13,14 The origin of such an effect remains elusive: the enhancement was ascribed to oxygen vacancy segregation at the grain boundary core according to theoretical investigations by Polfus et al. 15 Very recently, however, Börgers et al. have investigated fast mass transport along LSM dislocations, highlighting the role of space-charge effects. 16 The authors have also demonstrated that local oxygen kinetics in LSM can be ad hoc tailored by modifying the bulk stoichiometry. 17 Similar local effects on oxygen kinetics were found for La 0.8 Sr 0.2 Mn 1Àx Co x O 3 (LSMC) and for doped lanthanum chromite. 18,19 Notably, GB diffusivity and surface exchange coefficient in Sr-substituted thin lms of lanthanum chromite are remarkably larger (two and one orders of magnitude, respectively) than for LSM GBs. 20 While these results point out the potential of interface engineering for tuning oxygen kinetics and transport, the study of local mass transport presents a critical issue due to the impossibility of capturing such nanoscale effects with sufficient spatial resolution. Information obtained from diffusion proles acquired by SoA techniques based on isotope exchange depth proling (IEDP) is averaged over large areas (typically hundreds of mm 2 ) 21 and no information on the local chemical composition is usually accessible. These limitations prevent a complete understanding of diffusion phenomena and cause a considerable uncertainty for the retrieval of the kinetic parameters. 22 Very recently, we have demonstrated how atom probe tomography (APT) is able to overcome such shortcomings by providing a 3D-resolved reconstruction of atomic positions with nmresolution and isotopic sensitivity. 20 In our previous studies, isotope exchange of thin lms under controlled conditions allowed a direct observationand reliable quanticationof fast oxygen diffusion pathways at LSCr grain boundaries and LSM-ceria interfaces. 12,20 Independent work by Kaspar et al. recently presented a similar approach for the study of buried Fe 2 O 3 interfaces, conrming the broad relevance of the isotope exchange APT (IE-APT) approach. 23 In the present study, IE-APT was used for the rst-time visualizationand direct quanticationof previously reported enhancement of local oxygen transport in interfacedominated LSM. Fully dense nanocrystalline thin lms have been fabricated by pulsed laser deposition (PLD) for a subsequent ex situ exchange in an 18 O-enriched atmosphere at controlled temperature. Sample conical tips of the exchanged lms have been analyzed by APT providing a precise 3D chemical mapping and a direct visualization of nm-wide fast diffusion oxygen pathways. The experimental oxygen isotope fraction proles have been simulated by nite element modelling (FEM) for a precise and independent quantication of the oxygen kinetic parameters for bulk, GBs (self-diffusion coefficient, D*, and effective surface exchange coefficient, k*) and for the analysis of mass transport in sub-surface regions. Local cationic distribution, as retrieved by IE-APT, is correlated to the observed functionalities.

Results and discussion
Nanocrystalline LSM thin lms were deposited on top of ceriabuffered Al 2 O 3 (0001) substrates. Single-phase LSM with a columnar fully dense structure (average grain size of 27 AE 3 nm) was obtained aer deposition ( Fig. 1a and ESI Fig. 1 †). 17 The lms were rst annealed at 550 C in an 18 O-enriched atmosphere, allowing the tracer to diffuse into the layer, and then analyzed by APT for the measurement of the 18 O fraction. Fig. 1, panels b-d shows the results of the IE-APT analysis. A sketch of an exchanged APT tip containing a single grain is represented in Fig. 1b Fig. 1d, Z ¼ 20 nm). In the out-of-plane contour plot, one can clearly identify channels of higher oxygen isotope content (up to f( 18 O) z 7 Â 10 À2 ), which correspond to narrow and fast conducting regions and which can be spatially correlated to the structurally observed GBs. The horizontal separation between such regions is z30 nm, consistent with the grain size as retrieved by AFM and cross-sectional SEM ( Fig. 1a and ESI Fig. 1 †). A detailed analysis of the oxygen tracer fraction for bulk and GBs is reported later in the text. The in-plane contour (Fig. 1d) highlights the columnar shape of the 18 O accumulation GBs as shown in Fig. 1a and elsewhere. 17 Note that the apparent 18 O accumulation at the lateral edge of the tip should be interpreted as an experimental artifact likely due to surface damage/ contamination from the Ga focused ion beam preparation of the specimen. The region was conned to the edge of the analyzed specimen and it was observed to have high Ga, C, and H concentrations and greater overall background counts in the mass spectrum. The large eld of view obtained from these specimens, which was not observed previously, 12,20 is tentatively assigned to the specic lm-substrate combination and in particular to the very high evaporation eld of the alumina substrate. 24 FEM simulations have been employed in order to calculate the values of the oxygen transport parameters of LSM bulk and GB (k* and D*) by tting the APT reconstruction proles - Fig. 2. 2D model geometries were selected for accurately reproducing the out-of-plane IE-APT reconstructions (cf. Fig. 1a and ESI Fig. 2 †), which present grain interior areas (characterized by D * g and k * g ) separated by 1 nm-thick GBs (with kinetic parameters D * gb and k * gbfast GB core). 18,20 An additional parameter ðD * g;s Þ was introduced in the model for qualitatively describing the (slower) diffusion of the grain interior in the subsurface region, which accounts for the observed initial steep decrease of the tracer fraction for Z < 5 nm (Fig. 2b). This assumption is justi-ed on the basis of strong cationic deviations measured by APT cf. later in the textand of previous reports e.g. on Fe-doped SrTiO 3 , reporting slow subsurface oxygen diffusivity as a consequence of space-charge effects. 25 Please refer to ESI Notes 1 and 2 † for details on the FEM modelling, including sensitivity analysis. Fig. 2b shows the 2D contour of the FEM model aer tting of the experimental data (Fig. 2a) by systematic parameterization. The optimized FEM model is able to accurately describe the IE-APT results (apart from very local inhomogeneities) yielding: (i) fast oxygen incorporation and diffusivity at the GB (D * gb ¼ 2:1 Â 10 À14 cm 2 s À1 , k * gb ¼ 4:3 Â 10 À10 cm s À1 at 550 C); (ii) slower diffusivity and surface exchange rate for the bulk (D * g ¼ 1:0 Â 10 À16 cm 2 s À1 , k * g ¼ 4:5 Â 10 À12 cm s À1 ); (iii) reduced sub-surface diffusivity (D * g;s ¼ 1:5 Â 10 À17 cm 2 s À1note that the sub-surface region width is assumed as 5 nmcf. later in the text).
Unlike conventional tracing methods based on secondary ion mass spectrometry (SIMS)which offer only averaged data over hundreds of mm 2 -IE-APT allows extracting direct information on the local tracer fraction with nm-resolution in all directions (X-Y-Z). A segmented analysis of the data is therefore accessible for a separate study of the different sample areas as shown in Fig. 2c, in which light and dark green dotted lines correspond to 1D out-of-plane experimental proles along bulk and a selected GB, respectively. In-plane 1D proles of f( 18 O) are represented in Fig. 2d (proles extracted along the dotted lines in Fig. 2a). Using these data, f( 18 O) for the two phases can be independently t for a very precise quantication. The results highlight that isotopes incorporation in the bulk grain originates from lateral diffusion from the GBs (D * g ¼ 1:0 Â 10 À16 cm 2 s À1 ) giving rise to a concentration gradient in the X-direction (Fig. 2d). Conversely, vertical mass transport through the bulk is strongly suppressed. Finally, a local accumulation of oxygen isotope is conrmed in the sub-surface (cf. grain f( 18 O) prole in Fig. 2c, Z < 5 nm). This effect can be modelled by the introduction of the previously mentioned sub-surface diffusivity, D * g;s , which strongly improves the tting of the bulk tracer prole (while the quality of the GBs modelling remain largely unaffectedplease refer to ESI Fig. 3, † which includes two FEM ttings obtained for the same set of parameters, with and without D * g;s ). Effects stemming from trajectory aberration are known to limit the APT resolution to z2 nm at phase contacts, preventing a local study of the oxygen content in the grain boundary core. 26 In ESI Note 3, † we discuss an alternative FEM model based on the space-charge accumulation of oxygen vacancies (and strong reduction of oxygen diffusivity in the core). This behavior may originate from the segregation of the acceptor dopant (Sr 0 La ) or cationic vacancies (V 000 La or V 000 Mn ) in the core, giving rise to a negatively charged interface and promoting the accumulation of positively charged oxygen vacancies in the GB surroundings. We note that, in bulk LSM under oxidizing conditions, the concertation of cationic vacancies and acceptor dopant Sr is high, 27 while oxygen vacancies, which are typically expected to segregate at the GB core in ionic conductors, 28 are a minority defect. This interesting scenario was recently proposed by Börgers et al. to explain fast diffusion along dislocation in LSM. 16 The result of the simulations (ESI Fig. 8 †) show that also this model is able to adequately describe the f( 18 O) proles. 15,16 Importantly, the resulting kinetic parameters are very close to the ones obtained by considering fast core diffusivity (Fig. 2).
The resulting k* and D* are compared to literature data (stoichiometric 13 and B-site decient LSM 14 ) in Fig. 3: the values (measured here for T ¼ 550 C) are well in line with previous literature reports, conrming the quality of our approach based on IE-APT. The observed enhancement of both k * gb and D * gb is about two orders of magnitude with respect to the bulk. D * g;s is reduced by about one order of magnitude in comparison to D * g , yet within typical values for bulk LSM. Fig. 4 shows independent 2D contour plots of the different ionic species (La, Mn, Sr and O in Fig. 4a-d, respectively) and linescans along horizontal (Z ¼ 20 nm in Fig. 4f) and vertical directions (X ¼ 0 in Fig. 4g). As a rst general observation, important deviations from the nominal stoichiometry are present for A-site cations with Sr-content x ranging from $0.10 to $0.5 in La 1Àx Sr x MnO 3 , within a length scale z10 nm. (Please note that some uncertainty in the absolute atomic concentration may be introduced by the APT technique). 12,30 Fig. 4a (La contour) also highlights a strong modulation of La concentration (especially in the proximity of the fast oxygen diffusion pathways), which is complementary to Sr (cf. also Fig. 4c and f). Although such a relation is predictable (Sr being an A-site substitutional dopant), such a very broad non-stoichiometry at the nanoscale level is unexpected. It is noteworthy that this uneven distribution of A-site cations does not involve variations in the Mn distribution map within the resolution of the technique (Fig. 4b), con-rming a homogeneous presence of the ABO 3 perovskite phase (cf. also Fig. 1). Intriguingly, the variation of the A-site distribution is different for each GB (labelled as GB 1 and GB 2 ) and clear compositional differences can be observed between the two GBs. This is particularly clear in the 1D proles shown in Fig. 4e (crossing GB 1 and GB 2 ). Namely, GB 1 is characterized by a strong Sr accumulation (up to x $ 0.5) that is not observed for GB 2 (which is closer to stoichiometry). In both cases, Sr depletion occurs especially on one side of the boundary. As far as the total oxygen content is concerned ( 18 O + 16 O), a clear depletion is found in correspondence to the high Sr concentration areas (Fig. 4d), in agreement with a classical electroneutral situation 31 Oxygen diffusivity, however, seems not to be directly related to such strong compositional variations, as similar kinetic parameters characterize GB 1 and GB 2 (cf. Fig. 1). We ascribe such an apparent discrepancy to the different lengths scales under consideration (z10 nm for the observed stoichiometry changes vs. <3 nm for fast-conducting GB width), 16,17 alongside possible mobility effects. Note that the observed strong dopant segregation is expected to have a stronger impact on the electronic conductivity and magnetic properties of LSM. [32][33][34] Overall, such ndings (and similar observations previously reported by our group for nanocrystalline lanthanum chromite) 20 call attention to the need of accurate models which capture the different driving forces for a proper description of local chemistry and relevant cation non-stoichiometry in nondilute systems. A Poisson-Cahn approach, accounting for the electrostatic, but also structural and elastic contributions in nondiluted system, could in principle provide the rationale for the observed ion accumulation at the grain boundaries. 16,35,36 Improved models will be of utmost importance to understand and qualify the recently reported tendency of dopant de-mixing in MIEC materials. 37 Lastly, the out-of-plane ionic proles are analyzed in Fig. 4f (grain interior). A progressive decrease of the Mn content towards the free surface of the lm (Z ¼ 0) is highlighted, alongside a very strong accumulation of Sr and La in the subsurface region (Z < 10 nm). 38,39 Notably, the thickness of such a region, characterized by different stoichiometry, corresponds to the oxygen isotope accumulation area highlighted in Fig. 1 and can therefore be tentatively accounted for the local low D * g;s described previously. Note also that APT is able to retrieve local information on single grains and is therefore in principle unaffected by intergranular porosity/roughness for the assessment of the surface region.

Conclusion
A complete description of local oxygen mass transport properties for nanocrystalline LSM is reported by a novel approach consisting on isotope exchange depth proling combined with atom probe tomography. The analysis provides the rst direct visualization of the recently reported fast grain boundary oxygen diffusion in LSM, with nm-resolution in all directions. Experimental data include cationic and oxygen twodimensional distribution maps and have been t, for the case of isotopic oxygen, by an accurate nite element modelling of the microstructure. FEM, provided with an accurate independent quantication of oxygen mass transport properties for LSM at the bulk and grain boundary levels, revealed values of D * gb ¼ 2:1 Â 10 À14 cm 2 s À1 and k * gb ¼ 4:3 Â 10 À10 cm s À1 at 550 C for grain boundary diffusivity and surface exchange coefficient, respectively. These are in good agreement with previous gures obtained by conventional averaging techniques such as IEDP-SIMS. Local compositional maps of cationic distributions disclosed the presence of remarkable inhomogeneities in the Sr-dopant content (ranging from x ¼ 0.1 to x ¼ 0.5 in only few tens of nanometers), uniform Mn distribution and the presence of an A site-enriched subsurface region. The combined (and overlapping) quantication of cation, oxygen and isotope distributions allowed by IE-APT will likely become a common tool in the near future, giving rise to relevant data that will help developing robust models for explaining oxygen diffusion at the nanoscale or other relevant diffusive phenomena such as de-mixing effects in MIEC perovskites, which may strongly impact multiple applications like electrocatalysis, electrochemistry or magnetism.

Films fabrication
LSM thin lms were deposited by large-area PLD (PVD Systems -PLD 5000) using a 248 nm KrF excimer laser (Lambda Physics -COMPex PRO 205). The layers were deposited on Al 2 O 3 (0001) single crystal substrate (Crystec GmbH). A thin barrier layer of Ce 0.8 Sm 0.2 O 1.9 (SDC) was deposited before the LSM lm in order to avoid cationic intermixing at the interface. 40 Both layers were deposited at 700 C, under an oxygen pressure of 2.6 Â 10 À2 mbar, target-substrate distance of 95 mm, laser uency z1.2 J cm À2 and 5 Hz of laser frequency. The thickness of LSM and SDC layers deposited was z45 nm and z35 nm, respectively, as measured by spectroscopy ellipsometry (UVISEL, Horiba scientic).
Structural characterization X-ray diffraction was carried out using a Bruker D8 Advance diffractometer with a Cu K a radiation source in a 2q/u conguration (1.5 offset), step size 0.01 and counting time 1 s per step. Atomic force microscopy was carried out in a Park System and analyzed by Gwyddion soware. Scanning electron microscopy was carried out on a Zeiss Auriga Equipment (InLens detector).

Oxygen exchange
Aer cleaning the LSM thin lm sample surfaces by pure ethanol and acetone, an annealing in 200 mbar of pure oxygen N5 (99.999%) with 18 O 2 in the normal isotopic abundance was performed. Once the samples were cooled down at room temperature, the exchanged tube was evacuated and lled with an $90% 18 O 2 enriched gas (200 mbar). Following the exchange, the samples were quenched to room temperature. The nominal exchange temperature and time were 550 C and 1 h and 40 min, respectively.

Atom probe tomography
Atom probe tomography (APT) specimens were prepared using a li-out technique in an FEI Helios NanoLab 600i focused ion beam/scanning electron microscope (FIB/SEM). Specimens were mounted on TEM grids and hardware that allowed for TEM imaging (FEI Talos F200X) and analysis of the APT specimens. 41 Initial shaping was performed using a 30 kV Ga + ion beam voltage, with 2 kV used for nal sharpening.
APT (Cameca LEAP 4000X Si) was performed at 45.5 K using a 30 pJ laser energy and 500 kHz pulse rate. The ight path length was 90 mm and the ion detection rate was set to 5 ions per 1000 pulses, resulting in a bias range of 5000-7400 V during the data collection. Reconstructions were generated in Cameca's IVAS 3.6.18 soware using the TEM images of the specimens before and aer APT analysis (ESI Fig. 9 †) for setting the reconstruction parameters. 42 A systematic energy decit correction was employed to improve the mass spectral resolution. 43 Finite element modelling FEM simulations were performed by COMSOL Multiphysics to model the heterogeneous 18 O incorporation and diffusion in the LSM thin lms. The 2D geometry was modelled directly on the vertical cross section obtained by the IE-APT measurement (cf. Fig. 1 and ESI Notes 2 and 3 †). Two different diffusivity coefficients were considered for bulk and GB regions (D * b and D * gb ). The width of the GBs was xed to 1 nm, in accordance with previous works. 13,18,20 Oxygen incorporation was modelled by a convective-type boundary condition, employing two different surface exchange parameters for bulk and GB surfaces (k * b and k * gb ). To reproduce the steep decrease of tracer observed in the bulk subsurface, a different diffusivity coefficient was considered in the rst 5 nm ðD * b;s Þ. Fick's second law of diffusion was then solved by FEM method. The 2D results of the 18 O fraction obtained by the simulations were cut with the shape of the APT tip in order to offer a more direct comparison with the experimental data. Details on the FEM simulations including sensitivity analysis can be found in ESI Notes 1 and 2 † and in our previous works. 18,20

Conflicts of interest
There are no conicts of interest to declare.