Introduction

Present research in materials for solid oxide fuel cell (SOFC) technology is mainly focused in decreasing the operating temperature of the fuel cell to the intermediate range (500–700 °C) in order to avoid the materials’ thermomechanical degradation of the fuel cell components at higher operation temperatures (typically above 800 °C), which reduces the devices’ lifetime and obliges the use of expensive high-temperature-resistant materials, increasing the overall cost of the cells. This is particularly important when dealing with cell stacks where ceramic interconnect elements based on expensive chromium oxides or chromium-based metal alloys are required. The reduction of the working temperature diminishes the degradation and the interdiffusion between the different materials in the cell. Besides, it reduces the mechanical failure induced by the differences in thermal expansion coefficient of the different materials. However, it has a direct implication in slowing down the kinetics of the different thermally activated processes taking place in the cell, mainly oxygen surface exchange at the electrodes and bulk oxide ionic transport across the electrolyte. For this reason, it is necessary to explore new materials with increased performance or to develop new strategies with existing materials for enhancing their conductance, which may include reducing the thickness of the electrode/electrolyte current path or modifying the materials microstructure to induce fast transport paths along non-blocking interfaces or other type of defects. These different approaches will be discussed in detail in the following paragraphs.

In the last decades, there has been a progressive tendency towards the use of thinner SOFC elements, particularly in the case of the electrolyte, which is one of the main causes limiting the oxide ionic transport in the cell. In order to attain large power densities of the order of 0.1–1 W/cm2 at intermediate working temperatures, the implementation of methods to deposit thick films (from a few hundreds of microns down to almost 1 μm) has been required and progressively more complex thin-film technologies for thicknesses below 1 μm have been developed. Most current applications require large power generation, in the kilowatt and megawatt regime, making necessary the fabrication of SOFC devices with large effective surfaces, mostly stacks. However, there is a niche for lightweight low-power sources for devices requiring below 10 W, particularly for portable devices and remote small-scale sensors/actuators. For these devices, the fabrication of micro-solid oxide fuel cells (μSOFCs), based on similar thin-film technologies as those developed for microelectronics industry, is an emerging and highly active field of research that is attracting greater attention [1].

For the knowledge of the fundamental physicochemical properties of the materials involved in SOFC devices, it is essential to be able to separate those that are intrinsic to the materials from those that are caused by their microstructure or the presence of secondary phases or interfaces. This is sometimes a difficult task to achieve when dealing with polycrystalline ceramic samples and particularly if the films are prepared from methods that use additives and high-temperature post-treatments, in which there are too many parameters to control. For this reason, from a fundamental point of view, it is advisable to use test samples with more controlled microstructures and composition. This is generally the case of single crystals or thin films deposited from the vapour phase, either by physical or chemical methods. In most of the cases, it might be extremely difficult to grow single crystals of complex oxide materials, and the best approximation to obtain a material with similar properties to the single crystal is to grow epitaxial films.

In the following, we will only refer to oxide conducting materials for SOFCs, although in the last years there has been an increasing interest in oxide materials with protonic conductivity. The characteristics of those materials are sometimes very similar to those of oxide ionic conducting materials; for a comprehensive review on that topic, the reader is referred to [2].

Epitaxial thin films for SOFCs

The term epitaxy derives from the Greek roots epi, meaning ‘upon’, and taxis, meaning ‘arrangement, order’. Epitaxial growth refers to the continuation of the alignment of crystallographic atom positions in the single-crystal substrate into the single crystal film. More precisely, an interface between film and substrate crystals is epitaxial if atoms of the substrate material at the interface occupy natural lattice positions of the film material and vice versa [3]. Epitaxial films constitute an ideal frame for fundamental studies of the intrinsic material transport properties, offering, in the case of anisotropic materials, the possibility of independently measuring each direction’s component. In addition, they can also be used as models to study the properties variation by using substrate-induced strain.

Finally, this type of growth enables the deposition of artificial film heterostructures engineered at a nanoscale level, which can enhance the electrochemical performance of SOFC materials, as will be discussed in the next sections.

Deposition techniques for epitaxial thin films

Currently used techniques for the deposition of highly controlled epitaxial thin oxide films and multilayers include molecular beam epitaxy (MBE), pulsed laser deposition (PLD) [4], RF magnetron sputtering, e-beam evaporation, metal-organic chemical vapour deposition (MOCVD) and atomic layer deposition (ALD). It is not the aim of this work to describe in detail those techniques; an extended review on the different techniques can be found in [5].

During the deposition process, usually, the required material (or combination of materials) is supplied from a source (or different sources) and transported in vapour phase towards the substrate surface, where the deposition and further film growth take place. There are two main important requirements in order to obtain an epitaxial growth. First, the material has to be supplied to the substrate surface in atomic, molecular or ionic form with species consisting of a very few atoms, in order to guarantee a proper mixing of the constituents at a nanometre scale. This is easily achieved in vapour form produced from a solid source by thermal, arc, laser, e-beam evaporation or sputtering or from a liquid source by evaporation (pulsed-injection MOCVD) or spraying (spray pyrolysis). Then, the species are transported, mixed if necessary and adsorbed onto the surface by a physical or a chemical process. Second, there should be sufficient energy of the adsorbed species to diffuse along the surface in order to reach the equilibrium sites to form the crystal. In some cases, the incorporation into the solid is also mediated by a chemical reaction between the elements to form a different compound (CVD, ALD and reactive evaporation). Normally, the required energy is supplied thermally either by heating the substrates at temperatures from 400 to 1,000 °C or by activated processes making use of plasma sources, as in plasma-enhanced MOCVD, pulsed laser deposition or sputtering, techniques that allow the use of lower substrate temperatures while maintaining a high degree of crystallinity.

If the surface diffusion during the film deposition is limited by insufficient energy or by a too fast material supply rate, deposits with a low degree of crystallinity, amorphous or even porous films, are obtained. Still, in those cases, in order to activate the bulk diffusion, a post-annealing process at high enough temperatures (usually above the deposition temperature) may result in the rearrangement of the elements in the solid and may produce highly crystalline films.

One of the most interesting characteristics of vapour-deposited films is that the film growth generally takes place at conditions far from thermodynamic equilibrium, thus allowing the growth of metastable materials with unique microstructure, which may show different properties from the bulk materials. One example of these metastable structures is the material obtained in multilayer form by sequential deposition of two different compounds. In the extreme case of superlattices, by combining sequential growth of two different structures at elementary level (with control over atomic layer or unit cell thickness and perfectly flat interfaces), it is possible to obtain artificial structures with a new periodicity and, in some cases, to induce novel physical properties.

Substrate requirements

In order to grow an epitaxial film of a certain material with its atoms perfectly arranged on top of the atoms of the selected substrate, the first requirement is to have a good crystallographic lattice match between film and substrate for the chosen orientation (generally below 1% mismatch) [6]. If the film and substrate lattice parameters are exactly (or virtually) the same, the film is expected to grow unstressed, and therefore, the intrinsic material properties of the film can be measured. In this case, the film’s properties would be equivalent to those of a single crystal of the same composition (both the preparation and the characterisation of thin films being much simpler and unproblematic than those of single crystals). Alternatively, the choice of a substrate with a certain mismatch, i.e. with cell parameters either somewhat larger or smaller than those of the film, results in films grown under tensile or compressive strain, respectively. Figure 1 illustrates as example a perovskite epitaxial thin film of LSCF (a = 3.8753, b = 3.8769, and c = 3.8827 Å) grown on SrTiO3 (a = 3.905 Å) under tensile strain (Fig. 1a) and the same film grown under compressive strain (Fig. 1b) on a different substrate (NdGaO3, basal ab plane parameter a p = 3.866 Å). The implications of the epitaxial strain in the materials’ properties will be explained in the subsequent section. Finally, if the lattice parameters of the film and the substrate differ highly in all possible matching crystallographic orientations, it will not be possible to grow an epitaxial film, and the result will be either a preferentially oriented or a randomly oriented film.

Fig. 1
figure 1

Scheme of a perovskite epitaxial thin film of LSCF (a = 3.8753, b = 3.8769, c = 3.8827 Å) grown on SrTiO3 (a = 3.905 Å) under tensile strain (a) and on NdGaO3 (basal ab plane parameter a p = 3.866 Å) under compressive strain (b)

Furthermore, it has been proven that, by controlling the termination plane of the substrate surface structure at an atomic scale, the chemical composition of the terminating layer can be varied. Kumigashira et al. [7] reported on the fabrication of La0.6Sr0.4MnO3 (LSM) thin films in which the terminating layer was changed from B to A site by inserting one atomic layer of SrO between the LSM film and a TiO2-terminated SrTiO3 (001) substrate. This is expected to have an influence in the surface exchange properties of the material and might be crucial for the film growth of heterostructures with sharp interfaces [8, 9].

The second substrate requirement, in order to obtain a thermally stable film, is to have similar thermal expansion coefficients between film and substrate. Otherwise, when temperature changes take place (first from deposition temperature to room temperature and afterwards by subsequent cycling between room and working temperatures), cracks could appear on the film’s surface, as well as delamination or corrugation problems.

In addition, for a good chemical stability between film and substrate, the materials should be chemically inert at the film–substrate interface under the specific experimental conditions. Alternatively, if this condition is not fulfilled or if the mismatch with the substrate is too large, it is possible to use buffer interlayers of chemically inert materials, in which cell parameters should also be in good match with the film and the substrate.

Finally, depending on the thin films’ further use and characterisation, additional properties can be required. One should always confirm that there is no interference or interaction with the substrate’s properties. For example, if the intention is to selectively measure the electronic transport properties of the thin film, the substrate should be insulating, whereas in the case of characterising the electrochemical properties of the thin film when working as an electrode, the substrate should be oxygen conducting to act as electrolyte.

Epitaxial strain

In the case a certain film–substrate mismatch is present, different mechanisms, either elastic structural changes or plastic deformations mediated by the generation of defects, permit partial or total strain accommodation. These structural and compositional changes in the materials can induce, for example, variations in the electronic and ionic transport of the films. For instance, epitaxial films of the perovskite La0.8Sr0.2CoO3 (LSC) deposited on yttria-stabilised zirconia (YSZ) single crystals by PLD showed different electrode properties depending on the orientations. The LSC films on the YSZ (100) and (111) substrates showed the (110) orientation with different twin structures, while those on the YSZ (110) had (100) and (112) orientations [10]. Ideally, one could tune the electrolyte or electrode properties by using substrates with different cell parameters in order to optimise the performance of the SOFC. When studying strain-induced effects occurring at intermediate or high operation temperatures, one should always take into account the thermal expansion coefficients of both substrate and film, as the induced stress could be consequently reduced or enhanced.

There are some limitations for the deposition of strained films which are related to the elastic properties of the film material. The thicker the film, the higher is the elastic energy accumulated, proportional to the volume of the material. Above a certain threshold thickness, the material is not capable of maintaining its structure and tends to release energy either by generating defects (i.e. misfit dislocations, domains with different orientations and phase segregation) or by modifying its surface morphology (surface roughening and island formation). Those effects have to be taken into account and could be of interest for the enhancement of the materials performance, such as in the case of ionic transport along defects, as discussed in ‘Electronic transport’.

Transport anisotropy

In the case of anisotropic materials, such as layered mixed ionic–electronic conducting (MIEC) cathodes, it is particularly interesting to be able to grow epitaxial films of different orientations, which can be done by either using substrates with different cell parameters or by varying the deposition conditions of the thin film. This opens the possibility of evaluating anisotropic properties in different directions and of maximising the properties of a particular SOFC component (either the electrolyte or the electrode) by growing the material in a particular crystallographic direction. Figure 2 exemplifies the case of a double-perovskite material, GdBaCo2O5 + x , grown on a SrTiO3 substrate in three different directions: a-, b- and c-oriented. As this material is anisotropic, as will be discussed in detail in ‘Measurements’, the properties of the epitaxial thin film could be different for each of these three cases. This anisotropy has been proven by Shinomori et al. who grew a- and c-axis-oriented epitaxial thin films of layered nickelate La2 − x Sr x NiO4 on LaSrAlO4 (100) and (LaAlO3)0.3(SrAl0.5Ta0.5O3)0.7 (001) substrates (LSAT), respectively [11]. In the a-axis-oriented films, the resistivity and the optical spectra showed large anisotropy between the b and c axes, indicating the quasi-two-dimensional nature of the electronic structure in La2 − x Sr x NiO4. Another example is the low-temperature anisotropic oxygen diffusion reported by Inoue et al. in perovskite structure iron oxides [12]. Deposition on STO and LSAT yielded pseudo-tetragonal CaFeO2.5 epitaxially grown thin films oriented on the c p(b)-axis and CaFeO2.5 oriented on the a p-axis on LAO and LSAO substrates. The structural changes when single-crystalline CaFeO2.5 films were changed into CaFeO2 by low-temperature reductions with CaH2 were observed, implying that oxygen diffusion in the brownmillerite structure is highly anisotropic.

Fig. 2
figure 2

Scheme of GdBaCo2O5 + x (a = 3.881, b = 7.829, c = 7.542 Å [185]) double-perovskite epitaxial thin films grown on a SrTiO3 substrate with different orientations: a-axis (a), b-axis (b) and c-axis (c)

The methodology for the characterisation of anisotropic ionic/electronic properties is described in the next section.

Characterisation of ionic/electronic properties

In this section, we will focus on the characterisation of the most relevant properties for thin films for SOFC: charge transport, diffusion properties and surface exchange rates. For the description of more general techniques commonly used for chemical composition, structure, orientation and surface morphology characterisation, the reader is referred to [5].

Electronic transport

Electric conductivity measurements on thin films are normally performed either by using DC or AC techniques. Here, we outline the main configurations as well as the particular limitations and problems encountered when working with thin films. A more comprehensive description about these types of measurements can be found in a recent review by Guo [13].

DC measurements

DC measurements give information about the overall charge transport process and very often include a combination of the different mechanisms taking place in the film. Longitudinal measurements can be performed in a very simple two-point contact configuration (illustrated in Fig. 3a), by defining two parallel metal contacts (typically Ag, Au or Pt contacts evaporated through shadow masks or directly painted with metal colloids) on the surface of the film, forcing the current to flow along the film plane. Note that, in the four schemes in Fig. 3, for a better visualisation, the thin-film thickness L has been represented much larger than its actual dimensions; in reality, the lateral dimensions and the distances of the electrodes are much larger than L. For the transverse measurements, it is necessary to define metallic contacts on the top and on the bottom of the film, as shown in Fig. 3b. Since the films are supported on a substrate, for the bottom contact, either the substrate is conducting enough (Nb-substituted SrTiO3 single crystals are very often used providing low resistance, particularly at low temperatures) or a thin metallic bottom contact (such as Pt, SrRuO3 and LaNiO3) is deposited between the film and the substrate. For the growth of epitaxial films, in order to guarantee the structure coherence of the whole heterostructure, the bottom metallic electrode should also be epitaxial, offering a good film–substrate match.

Fig. 3
figure 3

Different electronic transport measurements configurations: longitudinal two-point contact configuration (a); transversal configuration (b); longitudinal four-point contact configuration with parallel contacts (c); and longitudinal four-point contacts with van der Pauw rectangular configuration (d). Please note that the thickness of the thin film (L) is grossly exaggerated when compared with the thickness of the substrate; in reality, the lateral dimensions and the distances of the electrodes are much larger than L

There are some restrictions to the use of DC measurements in thin films:

  1. 1.

    There should be no substantial charge accumulation in none of the interfaces within the material system or the contacts’ surface, preventing the building up of a capacitance. In the longitudinal configuration (Fig. 3a), as the film thickness (L) is orders of magnitude smaller than the contact length and distance, the geometrical factor makes the capacitance negligible in comparison to other stray capacitances in the experimental setup. However, in the case of transverse measurements (Fig. 3b), the capacitances could become important, particularly in the case of electronic transport in semiconducting oxide materials, and even larger in the case of electronic insulating electrolyte materials, as those typically used in SOFCs. In those cases, AC measurements are normally performed.

  2. 2.

    The conductance of the film material should be large enough in comparison to any other conductance in the transport path. In highly conducting materials, the contact resistance might be comparable to the film resistance. In those cases, a proper four-point contact, either with parallel contacts (Fig. 3c) or in a rectangular van der Pauw configuration (Fig. 3d), enables the elimination of the contribution of the resistance generated from the voltage measurement at the current contacts. On the other hand, in the case of highly resistive materials and particularly for very thin layers, the film resistance might become comparable to the substrate resistance. A proper choice of the substrate material and of the measurement conditions (temperature and gas atmosphere) is required to guarantee a homogeneous current flow through the film section.

  3. 3.

    In most of the materials used in SOFC, there are several types of charge carriers involved (electrons, holes and ions) in the transport phenomena. Particularly, in mixed ionic–electronic conductors used in SOFC electrodes, the ionic conductivity can become comparable to the electronic conductivity (n-type or p-type component). Still, in those cases, under certain temperature and oxygen partial pressure conditions, the total conductivity can be associated to the majority carriers. However, no information is obtained from the minority carriers, which very often play an important role in the electrochemical processes. Indirect information can be extracted from the extrapolation of the conductances to the regions where those carriers become predominant. Alternatively, the use of proper materials as electrodes may allow the selective blocking of the majority carriers, in order to reach a steady state in which the conductivity is only related to the minority carriers. For instance, pure oxide ionic conducting YSZ electrodes will block the electronic conductivity, while dense Pt electrodes are considered oxide ion blocking. On the other hand, Ag electrodes are considered reversible for electronic and oxygen transport. These are the ‘Hebb–Wagner’ methods, described in detail in [14].

Impedance spectroscopy (AC measurements)

AC measurements are generally performed on semiconductors and insulating materials when the samples present charge diffusion with characteristic non-negligible time responses and in processes in which there is a coexistence of several charge transport phenomena, as in SOFC, i.e. oxide ionic transport within the electrolyte grains or across grain boundaries, electronic transport in the electrodes, oxygen incorporation from the gas phase by reduction on the cathode or fuel oxidation at the anode. Each of these processes shows a characteristic time constant τ (related to its equivalent circuit, mostly a combination of a resistance R and a capacitance C in parallel, so τ = RC).The frequency response analysis of the impedance allows, in some cases, differentiation between those processes. Often, the measurements are limited by a high impedance value in a particular frequency range above the instrument input impedance. A proper patterning of the film geometry, either by defining a long and thin meander path (thus increasing the film resistance), by using interdigitated electrodes (reducing the film resistance and increasing its capacitance) or by using extremely small point contacts (increasing the film resistance and reducing its capacitance in transverse measurements), allows adjustment of the RC product to the measuring setup. Impedance analysis can become a quite complex task, and the information extracted from the raw data is very often dependent on the model used for the equivalent circuit associated to the system. For a more general description of the impedance analysis for SOFC materials, the reader is referred to [15].

Steady values of the total electric conductivity at different temperatures and pO2 give information about the main carriers involved and about the defect equilibria associated to a particular material. From this information, it is often possible to extract some characteristic intrinsic properties of the material, such as the activation energies for thermally activated charge diffusion processes, the apparent gap energy (in semiconductors) and the reduction and oxidation enthalpies [1618]. However, it is still difficult to separate the convolution between carrier density variations from the mobility of the carriers. In bulk materials, it is often possible to separate both conductivity components either by performing Hall measurements or by measuring thermoelectric power [1921]. Hall voltage induced by an applied magnetic field is proportional to the carrier mobility. Therefore, in materials with fairly small carrier mobility, such as band insulators, ionic conductors and even in electronic polaron semiconductors, the characterisation by this method is hampered by the small voltages generated. Thermoelectric power measurements are based on the characterisation of the small voltages generated by a temperature gradient induced in the sample between two electrodes. Unfortunately, in thin samples, the small dimensions, along with the presence of a substrate material, make it extremely difficult to extract any information from this type of measurements, and so far, there are no reports in the literature for thin films of typical SOFC materials.

Oxygen diffusion and surface exchange properties

Knowledge of the relevant transport parameters of the solid oxide materials involved is of fundamental importance in the development of SOFC devices. A variety of techniques have been developed, by which it is possible to determine, in addition to bulk transport coefficients, surface exchange constants. There are three fundamental experimental techniques usually applied to obtain transport coefficients: (1) the chemical experiment (electrical conductivity relaxation), in which the oxygen stoichiometry is changed; (2) the tracer experiment (isotope exchange depth profile method), in which the tracer distribution is changed; and (3) the electrical experiment (impedance spectroscopy) in which an outer electrical potential gradient is applied as from a driving force. It must be noticed that the three diffusion coefficients, D Chem, D * and D Q, and three rate constants, k Chem, k * and k Q, obtained from the three techniques, respectively, are intrinsically different and, as a general rule, cannot be directly compared [22].

Epitaxial films constitute ideal model materials to measure diffusion coefficients, as they typically are completely dense. Accordingly, in the case of the materials’ characterisation, the oxygen transport will have no influence on non-kinetic issues, such as particle morphology and connectivity, porosity and tortuosity, typical in porous electrodes.

Electrical conductivity relaxation

Monitoring of the DC conductivity changes when exposing a sample to an oxygen partial pressure step variation in the surrounding atmosphere (Fig. 4) may provide information about the kinetics of the different processes taking place in the sample [23, 24]. These processes are mainly the oxygen surface exchange with the atmosphere, either the in-take or out-take depending on the initial and final pO2, and the oxygen bulk diffusion. When dealing with very thin samples, the time necessary for the oxygen to diffuse through the thickness of the film can be considered negligible in comparison to the time required for the oxygen surface exchange. Therefore, the conductivity changes mainly reflect the surface exchange processes and are generally adjusted by a single exponential time dependence (provided the pO2 step change is small enough to associate a linear response to the conductivity change). The characteristic time is related to the surface exchange rate k for a given temperature and pO2 range. The limitations for this measurement are related to the rate at which the gas can be exchanged in the experimental setup in comparison to the surface exchange kinetics and are therefore often restricted to low temperatures. Alternatively, Tragut et al. [25] developed a method (conductivity relaxation in the frequency domain) that produces a modulation in the total gas pressure at varying frequencies by means of magnetic valves, which induces a periodic response in the conductivity of the sample. By monitoring the amplitude and the frequency response of the delayed signal, the authors are capable of extracting the surface exchange coefficient of the sample. This technique has been validated and used for thin films, obtaining values of the surface exchange coefficient up to 700 °C for epitaxial BSCF films [26].

Fig. 4
figure 4

Schematic of the ECR measurements of a thin film on a substrate. Please note that the thickness of the thin film (L) is grossly exaggerated when compared to the thickness of the substrate

Isotope exchange depth profile method

The oxygen isotope exchange technique consists of exchanging the ambient oxygen surrounding an oxide, which is usually an isotope mixture primarily made up of 16O, with a gas enriched with 18O (or 17O). The specimen is usually pre-treated in the oxygen partial pressure and temperature of the exchange measurement itself in order to establish thermodynamic equilibrium and prevent effects of chemical diffusion within the sample bulk. The consequence of a difference between the tracer content of the gas and that of the solid phase is a tracer exchange in the solid. The total oxygen content (16O + 18O) of the oxide remains constant during the exchange process, that is, the chemical composition is not altered. And so, this tracer diffusion is in fact a counterdiffusion of the two oxygen isotopes [27]. Data on these coefficients can be obtained from isotope exchange depth profiling by secondary ion mass spectroscopy after partial isotope exchange has taken place at the desired temperature and oxygen pressure. The tracer diffusion coefficient and the surface exchange coefficient are obtained by curve fitting of the measured profile to the appropriate equation, which is deduced from the solution to Fick’s second law with the appropriate initial and boundary conditions.

The development and validation of a new methodology for the determination of anisotropic tracer diffusion and surface exchange coefficients for epitaxial thin films by Burriel et al. [28] opened up the possibility of measuring a wide range of anisotropic materials, thin films and multilayers. The sample configuration and an outline of the 18O diffusion profiles for the traverse and longitudinal oxygen tracer transport measurements are shown in Fig. 5. To measure the diffusion along the traverse direction, the oxygen concentration profiles are analysed using the analytical solutions to the diffusion equation for a plane sheet model developed by Crank [29]. And to determine the oxygen transport along the longitudinal direction in the plane of the film, a dense and uniform thin film of a different material (such as Au) is required to protect the film surface and prevent oxygen exchange from the surface. A trench should then be defined in the surface of the protecting film penetrating down to the substrate. The lateral edge of the film is then opened to allow the exchange with the 18O-enriched exchange gas phase and thus ensuring the diffusion of the 18O species only along the longitudinal direction of the film. The longitudinal concentration profiles obtained are analysed using a semi-infinite plane solution of the diffusion equation [28, 29].

Fig. 5
figure 5

Sample configuration for traverse a and longitudinal b oxygen tracer transport measurements. Reproduced with permission from [28] Royal Society of Chemistry © 2008

Impedance spectroscopy

Electrochemical impedance spectroscopy is frequently used to determine the diffusion, kinetic and thermodynamic properties associated with oxygen transport in mixed conducting nonstoichiometric oxides. This technique, as has already been mentioned in ‘Substrate requirements’, allows independent measurement of the various component impedances if a satisfactory model for deconvoluting the relevant time constants can be developed. First, the distinctive features observed in the impedance spectra have to be assigned to the contributions from the ionic conduction of the electrolyte, oxide ion transfer across the electrode/electrolyte interface, and the oxygen exchange on the film surface. Then, an equivalent circuit model is used to analyse the impedance results, from which the surface chemical exchange coefficients can be derived. Several examples in which this technique has been used to extract the surface exchange coefficients of epitaxial thin films can be found in the literature [10, 3032].

Recent developments in epitaxial SOFC materials

In the following paragraphs, we will show the recent developments achieved in the field of SOFC materials through the use of thin films, mainly in the form of epitaxial layers and heterostructures.

Thin-film electrolytes

One of the key points to increase the performance of solid oxide fuel cells has focused on the reduction of the electrolyte area-specific resistance. Obviously, the search for materials with improved ionic conductivity has been one of the most extended approaches. In this case, composition variations of existing well-known materials, such as ZrO2 or CeO2-based fluorites, by means of extrinsic substitution, and the development of new materials such as LaGaO3-based perovskites, BiMeVOx aurivillius phases, apatites, etc., have been the subject of extended reviews [3245]. But for real operation of SOFC devices, a more integral approach has to be adopted, taking into account other aspects, such as the chemical and thermomechanical compatibility with different types of electrodes. This aspect has been the subject of a major activity in the field [38, 4661].

Other studies have pointed towards a geometrical approach which is to reduce the electrolyte thickness and to increase its surface area. While the second is still in its infancy, there have been some attempts to produce corrugated YSZ electrolyte films by means of lithographic techniques [62], and some authors pointed out the possibility of using nanostructured templates, such as nanopillars or nanopores, to build up large specific area films. The reduction of the electrolyte thickness is an obvious way to increase the oxygen conductance, provided the material properties remain the same as in bulk material.

Given the multicomponent nature of the SOFC devices, once the oxide ion conductance at the electrolyte has reached a certain value, it might be that a further reduction of its thickness does not increase the performance of the cell, as there could be other steps in the overall process limiting the performance. Some authors have established a target value of area-specific resistance at about 0.15 Ω cm2 [63] for the electrolyte, below which a further reduction is supposed not to be effective, provided standard values of the conductances of the rest of the components. This resistance could be attained for a film of 1 μm at temperatures of about 500 °C for YSZ and of 300 °C for Gd-doped ceria, assuming bulk conductivity values. However, considerable improvements can be expected for all the components, which make further increase in the electrolyte conductance of great interest, i.e. further thickness reduction towards the sub-micron scale.

Although there have been large improvements in some of the large-scale fabrication techniques generally used for depositing electrolyte films, such as tape casting or screen printing, they have encountered some limitations when trying to deposit thin films with homogeneous thicknesses below 1 μm. To overcome these limitations, there is an increasing usage of well-developed vapour-phase techniques, as those described in ‘Deposition techniques for epitaxial thin films’, which allow depositing films down to the nanometre scale with reasonable uniformity.

Nanocrystalline films: grain size effect

As is generally the case in polycrystalline materials, the charge transport and mass diffusion mechanisms across the material can be substantially different in the bulk of the grains and in the grain boundaries. Obviously, the smaller the grain size, the larger is the ratio between grain surface and bulk effects. In nanocrystalline samples, the interface effects may dominate the transport mechanisms, which could significantly modify the performance of the electrolyte material. In some cases, these interface effects have generated an enhancement of the ionic conductivity, as in YSZ epitaxial ultrathin films [6467], as shown in Fig. 6. Kosacki et al. [66] reported two orders of magnitude enhancement in the film conductivity in epitaxial YSZ films of about 15-nm thickness deposited on MgO single-crystal substrates in comparison to bulk YSZ conductivity, along with a substantial reduction of the activation energy. A similar behaviour has also been reported for Gd-doped ceria films, as shown in Fig. 7. However, the reduction of the film thickness might also induce the enhancement of the electronic conductivity in an electrolyte material, as in the case of nanocrystalline CeO2 samples (also shown in Fig. 7) [16, 6873], turning the material into a mixed ionic–electronic conductor.

Fig. 6
figure 6

Arrhenius plot of the conductivity for thin epitaxial films deposited on MgO single crystals. Reproduced with permission from [66]. Copyright 2005, Elsevier

Fig. 7
figure 7

Electronic conductivity for undoped CeO2 (left) and ionic conductivity in Gd-doped CeO2 (right) epitaxial thin films. Reproduced with permission from [72]. Copyright 2002, Elsevier

Also, the grain boundary contribution to the electric transport may be different if the transport takes place along the interfaces or across them. In some cases, as in SrTiO3 polycrystalline samples, the boundaries could have a blocking effect when the current is forced to flow across the boundaries [74], whereas the transport is enhanced parallel to the interfaces [7577]. For the modelling of the whole charge transport mechanism, the combination of all contributions has to be taken into account, which is generally attained by using the ‘brick-layer model’ [7781].

Nanocrystalline ceramic bulk materials are generally prepared from fine nanopowders, or via sol–gel processes. In those cases, to attain densification with a reasonable connectivity between grains, it is necessary to use severe high-temperature post-annealing steps of the order of 900–1,300 °C. This may produce certain instability of the metastable nanostructure, inducing grain growth [82, 83] or impurity segregation towards the grain boundaries (as in the typical case of Si impurities [84]) where they accumulate, producing clear blocking effects [74]. An important difference between nanocrystalline materials prepared from the standard solid-state reaction and materials deposited in the form of films by chemical or physical vapour deposition methods (MOCVD, ALD, e-beam evaporation, sputtering, pulsed laser deposition, etc.) is that from the vapour techniques it is possible to attain dense nanocrystalline microstructures at the deposition temperatures (generally well below 900 °C) without the need of further annealing steps, minimising therefore possible impurity segregation towards the grain boundaries. Besides, under some conditions, the use of vapour-deposited films induces the growth of preferentially oriented grains (in a columnar structure) which would generate certain anisotropy in the grain boundaries (aligned across the thickness of the film) and could have a favourable effect either directly in the charge transport mechanism or in the stabilisation of the nanostructured domains.

Combination of insulating–conducting materials/space charge effects

One of the most interesting approaches to enhance the ionic conductivity of electrolyte materials is based on the space charge effects generated at the interface between two different materials [13, 85]. At steady-state conditions, the different electrochemical potentials of the two materials in contact would balance by interdiffusion of charge carriers, thus modifying their equilibrium bulk charge concentration at the interface area with an extent of the order of the Debye length of the material [86]. The values could go from only a few nanometres to several tenths of nanometres, depending on the materials involved. If the composite material consists of nanometric-size crystal domains, then the space charge effects may prevail over the bulk behaviour. In some cases, as in BaF2/CaF2 multilayered heterostructures obtained by MBE [87], the concentration of the dominant ionic mobile species (fluoride ion interstitials and fluoride vacancies) could be largely depleted and increased, respectively, in the space charge region of BaF2 material in contact with the CaF2. This depletion would cause an inversion of the majority carriers, therefore inducing an increase in conductivity by several orders of magnitude along the direction parallel to the interface region, in proportion to the density of interfaces. It is also foreseen that a further reduction of the layer thickness would also make it possible to increase the charge transport across the layered structure [88]. A similar effect occurred on the ionic conductivity of lithium iodide–aluminium oxide composite solid electrolyte associated to the space charge regions [89].

This is possible in very diluted systems with high carrier mobility but low intrinsic concentration (as in BaF2 where F is about 1017 cm-3 [13]). However, in oxide conducting materials such as disordered oxygen vacant fluorites or perovskites, in which the carrier concentrations are of the order of 1020–21 cm−3(that is a few per cent per formula unit), a further increase in the density of vacancies is very limited (hardly reaching one order of magnitude). Besides, it would readily lead to defect association or clustering, therefore limiting their mobility [9092], as has been proven by aliovalent substitution in bulk material [93]. In some cases, the combination of an insulating material, such as Al2O3, with Gd-doped ceria electrolyte material in the form of a nanocomposite has been proven to prevent the building up of predominant electronic conductivity by acting as a trap for the generated electrons produced by the reduction of Ce [94].

Multilayers of ionic conducting oxide materials

One of the first evidences of interfacial effects between two different oxides having an important role in enhancing the oxide ionic conductivity was reported by Azad et al. [95] in gadolinia-doped ceria and zirconia multilayers deposited by MBE on Al2O3 (0001) substrates. A cross section of the heterostructure is shown in Fig. 8. In such multilayers, the oxygen ion conductivity was increased about one order of magnitude at 350 °C, in comparison to bulk materials. Besides, the conductivity enhancement was proportional to the number of interfaces, which evidenced the interface effect. However, the short Debye length in these oxides ruled out the possibility of space charge effects, and the authors therefore suggested the strain or the presence of extended defects as a possible explanation for the conductivity enhancement.

Fig. 8
figure 8

TEM image of the cross section of an ionic conducting Gd-doped CeO2/ZrO2 multilayer deposited on Al2O3(0001) single-crystal substrate. Reproduced with permission from [95]. Copyright 2005, American Institute of Physics

An even greater enhancement of about two orders of magnitude in the electrical conductivity was also observed by Kosacki et al. in highly textured (30 nm) CeO2/(20 nm) 20% Sm-doped CeO2 multilayers [96] with a progressive reduction of the activation energy with the number of interfaces from 0.63 eV (for bulk doped material) to 0.46 eV (for a multilayer with 400 periods). However, it was never fully proven that the conductivity enhancement was purely ionic or that the electronic conductivity became predominant due to a reduced enthalpy for oxygen vacancy formation induced by the nanometric CeO2 layer thickness, in the same way as reported by Suzuki et al. for pure CeO2 nanocrystalline films [72]. A very recent work reported by Perkins et al. on the same type of CeO2/(20 nm) 15% Sm-doped CeO2 epitaxial multilayers deposited on MgO (100) substrates by PLD [97] showed ionic conductivities values very similar to those corresponding to the bulk material. In this case, in order to relax the epitaxial strain and thus eliminate that contribution to the conductivity, the films had been deliberately grown onto a buffer layer. Therefore, this last work points towards the role of strain or of the defects generated in the films associated to the epitaxial strain, rather than to the space charge effects, as the main cause for the ionic or electronic conductivity enhancement in this combination of materials. In a similar way, a recent report by Sanna et al. [98] showed an increase of about one order of magnitude in the conductivity of epitaxial Sm-doped CeO2/YSZ strained multilayers obtained by PLD on MgO (100) single-crystal substrates, supporting this assumption. The study of the pO2 dependence in Fig. 9 evidences not only the enhancement of both the ionic conductivity at high pO2 values, but also of the n-type electronic conductivity at low pO2 values, in direct relation to the number of interfaces in the heterostructure.

Fig. 9
figure 9

Conductivity dependence on oxygen partial pressure in epitaxial Sm-doped CeO2/YSZ heterostructures. Reproduced with permission from [98] © 2010 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Strain effects: effect on mobility

In oxide conducting materials, the increase of carrier concentration has serious limitations. However, the low carrier mobilities could be enhanced by some subtle structural modifications, which would induce variations in the interatomic distances, reducing the energy barrier for oxygen hopping along the diffusion paths. This enhancement has already been demonstrated by modifying the radius of homovalent substitutional cations [99]. In a similar way, it has also been demonstrated that, in thin films submitted to large biaxial stress (of the order of tenths of GPa) generated by the epitaxial growth onto a single-crystal substrate with different cell parameter, the strained structure may induce variations in the ionic mobility. However, depending on the elasticity parameters of the material, a strain larger than a few percent cannot be maintained. The accumulated elastic energy is generally released by the formation of dislocations (or other type of defects, as has been discussed in a previous paragraph), and the structure recovers its equilibrium values. In fluorites such as YSZ or CeO2 combined with different types of insulating oxides (Lu2O3, Sc2O3, Y2O3, Al2O3 and MgO) in the form of epitaxial multilayers grown by PLD, it has been demonstrated that the formation of a coherent interface between structures with different cell parameters induces a variation of the ionic mobility. The induced biaxial strain, either tensile (about +3% mismatch in YSZ/Y2O3) or compressive (−4% mismatch in YSZ/Sc2O3) in a region with a certain extent across the interface, would increase or reduce the interface conductivity, respectively, when compared with better matching structures (about 1% mismatch in YSZ/Lu2O3), where the YSZ ionic conductivity approaches that of the bulk material. This has been shown in several studies by J. Janek’s group from Giessen Univ. [100102], and those results are depicted in Fig. 10. A simple model just taking into account the elastic properties of YSZ, relating the volume ratio of the strained material with a volume increase and with a reduction of the activation energy for the oxygen vacancy migration, seems to consistently account for the observed effect.

Fig. 10
figure 10

(Top) TEM cross section of a 50× (YSZ/Sc2O3) multilayer deposited by PLD on an Al2O3(0001) single crystal. (Bottom) Film conductivity deviation from bulk behaviour for different heterostructures combining YSZ and RE2O3 (RE = Y, Sc and Lu). Reproduced with permission from [102] Royal Society of Chemistry © 2010

However, some authors point to the possibility that the generation of a large density of defects, such as dislocations in semi-coherent heterostructures or boundaries between columnar grains, may produce fast ionic diffusion paths, as attributed in the case of CSZ/Al2O3 multilayers reported by Peters et al., where about two orders of magnitude increase in the conductivity were observed [103]. Similarly, a more recent report by Sillassen et al. on YSZ epitaxial single layers deposited by sputtering on MgO (110) substrate [104] (with a very large mismatch of about +18%) showed an increase in the conductivity of about three orders of magnitude. The absence of grain boundaries made the authors associate the ionic conductivity to the presence of a dense array of misfit dislocations generated, accounting for the large mismatch.

Other studies on Gd-doped CeO2 epitaxial layers deposited by PLD on MgO (100) reported by Chen et al., growing on a cube-on-cube arrangement with an extremely large mismatch of about 28% and therefore expected to have a high density of misfit dislocations, did not show any particular enhancement of the ionic conductivity with respect to the bulk material [105]. These results demonstrate that not only the presence of dislocations but also the characteristics of those defects may be effective in producing fast oxide ionic diffusion paths.

The most astonishing results were reported by Garcia-Barriocanal et al. in STO (100)/YSZ (100) epitaxial multilayers deposited by magnetron sputtering [106] in which an enhancement in the ionic conductivity of about eight orders of magnitude was observed. The authors attributed the enhancement to a particular opened structure of the fluorite induced by the large tensile strain of YSZ at the interface. Although there have been some theoretical studies by first principles supporting this idea [107], these results have raised certain controversy within the solid-state ionics community about whether the conductivity enhancement in YSZ/STO system has a pure ionic character [108, 109] or whether it is related to an electronic conductivity induced by doping the STO substrate with acceptor impurities [108]. Another study on YSZ/STO multilayers deposited by PLD reported by Cavallaro et al. [110] has given a similar conductivity enhancement of several orders of magnitude. However, p-type conductivity dependence on pO2 with exponent of about −1/4, along with the measurement of no enhancement in the oxygen diffusion along the interface direction by means of isotopic oxygen exchange depth profile experiments, has evidenced the electronic nature of the transport mechanism in such system and related it to the possible formation of a highly conducting perovskite-type phase by interdiffusion at the interface.

Thin-film MIEC cathodes

In addition to the general stability and compatibility requirements of any SOFC material, those used as cathode electrodes need to have high electronic conductivity under oxidising conditions and a good catalytic activity toward promoting oxygen dissociation [111]. Current research in the field is mainly being directed toward materials with mixed ionic–electronic transport properties (MIEC), whose use would permit lowering the operation temperature of the cell to the range of 500–700 °C. The state-of-the-art cathode used at higher temperatures, the perovskite LSM, has a limited performance at the intermediate operation temperatures, and therefore substitute materials with improved properties should be identified. By using materials with high ionic conductivity in addition to the electronic contribution, the triple-phase boundary zone is effectively broadened to include the entire surface of the MIEC and, consequently, the polarisation resistance can be lowered.

Materials with the perovskite or related structures dominate the field, mainly including three families of oxides: single perovskites, Ruddlesden–Popper (RP) phases and layered double-perovskite phases. For a detailed description on recent advances on new cathode materials, the reader is referred to the comprehensive reviews by Orera and Slater [111], Jacobson [112], Sun et al. [113] and Tarancón et al. [114]. There have been some recent attempts to fabricate composite cermet cathodes in the form of thin films by combining pure ionic and electronic conducting materials, such as the vertically aligned LSCO/CGO nanocomposite films deposited by PLD by Yoon et al. [115], which have been already successfully tested in symmetric cells. Figure 11 shows the microstructure of such composite electrodes.

Fig. 11
figure 11

Cross section of PLD deposited LSCO/CGO vertically ordered nanocomposite cathodes. Low and high magnification (top and bottom, respectively). Reproduced with permission from [115] © 2009, John Wiley and Sons

Dense thin films have been extensively used to prepare novel electrode compositions and structures for use in fundamental reaction mechanism and electrochemical measurements [116134] and, as they are optimal materials with a well-defined structure, as model materials for the study of the oxygen surface exchange and bulk diffusion properties of these mixed conductors [119, 135, 136]. Otherwise, in the case of porous bulk samples, the true values of the surface exchange coefficient and diffusivity are actually masked, as they are influenced by non-kinetic material properties (such as density, porosity, etc.). The information extracted from these ‘model materials’ is key in the design of optimum electrode morphologies. Also, as pointed out by Beckel et al.[1], thin films can also be utilised to increase the performance of SOFC cathodes in miniaturised systems (μSOFC). Examples and details of thin-film cathodes for micro-solid oxide fuel cells can be found in the authors’ review article. To date, only a couple of examples of half cells with an epitaxial cathode film on an epitaxial electrolyte substrate have been reported. In such a configuration, the cathode material can be grown to take advantage of its maximum transport direction and, ideally, could be strain-tuned to further boost its MIEC properties.

Perovskite-type epitaxial films

The number of studies on epitaxial thin films of potential cathode materials is still rather scarce, and these have been mainly been carried out for perovskite-type materials [10, 26, 32, 137139].

For most of the perovskite materials used such as cathodes in SOFCs, the A site cation of the ABO3 structure is a mixture of rare and alkaline earths (such as La and Sr, Ca or Ba), while the B-site cation is a reducible transition metal such as Mn, Fe, Co or Ni (or a mixture thereof) [113]. Consequently, in most cases, the B-site cation usually provides a redox catalytic mechanism. A large and stable number of oxygen ion vacancies can be introduced at the operating conditions of the fuel cell by varying the A- and B-site cations and hence increasing significantly the bulk ionic oxygen transport.

Epitaxial La0.5Sr0.5CoO3 − δ film formation was first reported on the LaAlO3 substrate by Chen et al. [137], who determined its oxygen surface exchange coefficient from electrical conductivity relaxation measurements. The measured values of k chem were very low compared with the values for bulk polycrystal materials, consistent with the observation in literature for LSCO single crystal. Contrary to these results, O’ et al. have very recently published a markedly increased oxygen exchange by up to two orders of magnitude relative to the bulk for La0.8Sr0.2CoO3 − δ /GDC/YSZ (001) [139]. They attribute this behaviour to increased oxygen vacancy concentrations in the films, in which physical origin is not clearly understood. They suggest that cation defects/vacancies and/or strains are more likely to occur near the epitaxial LSC/GDC interface, which may stabilise more oxygen vacancies and thus enhance the oxygen surface kinetics. This constitutes the first report in which epitaxial oxide films exhibit greatly enhanced surface kinetics and, although a better understanding is highly required, the findings illustrate the potential to promote oxygen activity on selected heterostructured oxide films and provide new strategies to design highly active catalysts for applications in solid oxide fuel cells.

In two previous works, Mori et al. deposited epitaxial La0.8Sr0.2CoO3 − δ, La0.8Sr0.2MnO3 (LSM) and Y0.5Ca0.5MnO3 (YCM) thin films with the perovskite structure on YSZ substrates of different orientations and studied their electrochemical properties [10, 31]. The film orientations depended on the plane cut of YSZ substrate. Differences in the resistivity and activation energies were found to depend on the lattice misfit, surface structure and surface morphology, but further studies should be carried out to clarify the relationship between the surface structure and the electrochemical properties. LSM epitaxial films have also been used by Fister et al. to study strontium surface segregation using in situ synchrotron measurements of total reflection X-ray fluorescence [138]. 001-oriented La0.7Sr0.3MnO3 thin films were grown by PLD on DyScO3 (101) and NdGaO3 (101) substrates and measured over a wide range of temperatures and oxygen partial pressures. The strontium surface concentration was observed to increase with decreasing pO2, suggesting that the surface oxygen vacancy concentration plays a significant role in controlling the degree of segregation. In contrast, the La0.7Sr0.3MnO3 film thickness and epitaxial strain state had little impact on segregation behaviour.

Another perovskite material which presents improved properties when prepared as high-quality epitaxial thin films is Ba0.5Sr0.5Co0.8Fe0.2O3 − δ (BSCF). The electrical conductivity of the thin films prepared by PLD on single-crystal NdGaO3 (110) showed different activation energies and higher general conductivity [26] when compared to polycrystalline bulk values. In this case, the measured surface exchange coefficients are in the same range of magnitude as those measured for polycrystalline BSCF samples.

Ruddlesden–Popper phases

The homologous series of compounds with the general formula A n + 1 M n O3n + 1 that are structurally similar to the titanates, Srn + 1Ti n O3n + 1, reported by Ruddlesden and Popper in 1958 [140] are referred to as RP phases. The structure of the RP phases is made up of n consecutive perovskite layers (AMO3) n , alternating with rock-salt layers (AO), along the crystallographic c-axis direction, so their formula can be represented by (AO)(AMO3) n , where n represents the number of connected layers of vertex sharing MO6 octahedra. Of particular interest in the case materials for SOFC cathodes are the n = 1 members of the family, particularly those based on La2NiO4 + δ , which can accommodate oxygen interstitials in the rock-salt layers, giving as result fast ion conduction and hence potential application at intermediate temperatures. Given their layered structure, the physical properties of the n = 1, 2 and 3 phases are expected to be highly anisotropic.

Within the RP family and motivated by their superconductivity properties, Shinomori et al. [11] were the first to report on the synthesis of epitaxial thin films of Sr-doped lanthanum nickelate (La2 − x Sr x NiO4) on two different substrates and to measure their anisotropic electronic properties at low temperatures, as has been previously explained in ‘Transport anisotropy’. The synthesis of undoped c-axis-oriented La2NiO4 + δ epitaxial thin films has been reported by two different groups: by Kim et al., who deposited the films on LaAlO3 (001) substrates using PLD [23], and Burriel et al., who deposited the films on SrTiO3 (100) and NdGaO3 (110) substrates by pulsed-injection metal-organic chemical vapour deposition [141, 142]. In this second case, the electrical conductivity along the ab plane increased by decreasing the film thickness, reaching values of 475 S/cm for 33-nm-thick films [142, 143], higher than the values reported for bulk samples and even for single crystals. An important role of the absolute strain on the transport properties of this material was suggested and related to the structural evolution observed from a strained structure to a more relaxed one upon thickness increase. A complementary work from Burriel et al. [28] confirmed an anisotropic behaviour of the oxygen diffusion and surface exchange of the La2NiO4 + δ epitaxial thin films, showing two to three orders of magnitude larger values along the ab plane than along the c-axis (as shown in Fig. 12). For this purpose, the authors developed and validated a new methodology for measuring the surface exchange and oxygen diffusion properties of thin films in two perpendicular directions by the isotope exchange and depth profiling (IEDP) method, as has been explained in ‘Epitaxial strain’. Another finding of this work was the increase of the diffusion coefficients (both along the c-axis and along the ab plane) with film thickness, i.e. as the strain in the film is released. Contrary, the thickness did not have a direct effect on the surface exchange coefficients.

Fig. 12
figure 12

Arrhenius plot of the c-axis diffusivity in the zone close to the film surface (D *c1 ) and of the ab plane diffusivity (D *ab ) for La2NiO4 films deposited on STO and on NGO and comparison with the literature data for La2NiO4 single crystal and dense ceramics. Reproduced with permission from [28] Royal Society of Chemistry © 2008

In an attempt to obtain thin films of the higher members of the Ruddlesden–Popper series (Lan + 1Ni n O3n + 1, n > 1), Burriel et al. [144] varied and controlled the La/Ni stoichiometry of the films. In addition to epitaxial La2NiO4 (n = 1) and LaNiO3 (n = ∞) pure phases, for intermediate values of La/Ni composition, films with a microstructure consisting of a disordered stacking of nanodomains with progressively increasing average n value were obtained. Similar to the bulk samples, the electronic conductivity of the nanostructured films increased with the measured average n value, opening the possibility of preparing new mixed ionic and electronic conducting nanocomposite thin films with tailored properties, although no information about ionic conductivity in those phases has been reported to date.

Despite a high epitaxial mismatch of 9.98%, Yamada et al. [30] were able to deposit very thin (110)-oriented epitaxial layers (approximately 14 nm) of Nd2NiO4 + δ on a (100) YSZ single crystal by PLD, obtaining the first reported heteroepitaxial SOFC system. The authors claim that this K2NiF4-type structure might have some inherent flexibility that allows epitaxial growth despite its large mismatch. By impedance spectroscopy, the authors measured a large activation energy for oxygen diffusion along the ab plane, which they think could be related to the shrinkage of the bottleneck for the interstitial oxide ion diffusion, resulting from the significantly small measured c (in-plane) and from an expanded a (out-of-plane) cell parameter for the Nd2NiO4 film. This is an example of how the cathode properties can be optimised by preparing highly textured or epitaxial thin films with the planes or channels of maximum ionic transport aligned with the oxygen diffusion direction. This is especially convenient in the case of micro-fuel cells in which the cathode is a thin-film layer.

Double perovskites and layered ferrites

Ordered double perovskites with general formula AAB 2O5 + x , where A is a rare earth, A′ an alkaline earth and B either Co or Mn (GdBaCo2O5 + δ structure shown as example in Fig. 2), have been recently proposed as the next-generation cathode materials for intermediate- to low-temperature operation due to their high electronic conductivity and enhanced oxygen surface exchange rate and transport properties [114]. These layered cobaltites are being actively studied, not only due to their mixed conducting properties but also due to their magnetic properties, having shown phenomena such as a metal–insulator transition or colossal magnetoresistance effect [145]. The properties of these compounds depend strongly on the oxygen content and oxygen vacancy ordering, as well as on the ordering of the A and A′ cations. The structure ordering shown in Fig. 2, in which rare earth and alkali-earth ions occupy alternate (001) layers and oxygen vacancies are mainly located in the rare earth planes forming channels along the a-axis, occurs for each compound for a certain range of oxygen stoichiometry and is thought to be responsible for the high oxygen diffusivity values.

As these materials can be prepared with A site cation-ordered or cation-disordered structure at the same composition, their growth as epitaxial films has proven to be complicated. In addition, due to the defined oxygen content window at which the ordered oxygen vacancies appear, along with the peculiarities of vapour-phase growth, the selection of the appropriate substrate and deposition conditions has to be done very carefully in order to obtain epitaxial films of one sole domain orientation. A very nice example of how the preparation conditions can affect the film microstructure and properties of the films can be found in the work done by Kasper et al.[146] for epitaxial TbBaCo2O5 + x films. The authors managed to grow c-axis-oriented epitaxial films on SrTiO3 (STO) and (La, Sr)(Al, Ta)O3 (LSAT) substrates, while a mixture of c- and a(b)-oriented crystallites or c-oriented films with a large ‘out-of-plane’ mosaic was always obtained on LaSrAlO4 and LaSrGaO4 substrates, independent of substrate temperature and oxygen pressure during deposition. They also found that the phase with perovskite structure only developed if the oxygen pressure in the chamber during deposition was larger than 5 × 10 −3 mbar and that the superstructural reflections showing the ordering of the cations only appeared in films grown at substrate temperatures larger than 897 °C. Additionally, oxygen vacancy ordering was found for identical deposition conditions only if the cooling rate after deposition was very low (1 K/min).

The first successful attempt on the growth of double perovskites epitaxial films was carried out by Kim et al. [147] and Yuan et al. [148], who synthesised PrBaCo2O5 + x (PBCO) thin films by PLD on (001) SrTiO3, finding the coexistence of two domain structures in the films [147, 148]. The authors measured the oxygen exchange kinetics of thin films by electrical conductivity relaxation (ECR) and by oxygen IEDP, revealing high electronic conductivity and rapid surface exchange kinetics. The same group of authors have very recently reported on the epitaxial growth of LaBaCo2O5 + x thin films on (001) LaAlO3 single-crystal substrates by PLD [149, 150]. The films were claimed to be c-axis-oriented with very weak superstructure ordering reflections appearing along this axis, which could be either due to cation or oxygen vacancy ordering. After a treatment in hydrogen, the epitaxial nature was maintained while the superlattice peaks disappeared. The chemical dynamic studies on the films revealed a drastic change in resistance with a change of redox environment, within short response times, suggesting an extraordinary sensitivity to reducing oxidising environment and an exceedingly fast surface exchange rate. The results indicate that, at low oxygen partial pressure, the extension of oxygen deficiency is an essential factor to the high-temperature physical properties of LaBaCo2O5 + x and suggest its potential application as a cathode material in intermediate-temperature SOFCs or as a chemical sensor device for reducing environments at high temperature.

The possibility of controlling the growth on a unit cell level for another double cobaltite compound, namely NdBaCo2O5 + δ , has been achieved by Grygiel et al. [151] for epitaxial films on (001) SrTiO3 substrates (Fig. 13 shows a high-resolution transmission electron microscopy (TEM) image of an ordered NdBaCo2O5 + δ film exhibiting doubling of the out-of-plane lattice parameter). By adjusting the growth kinetics via control of temperature and laser energy, cation-ordered units (with double cell parameter) or cation-disordered units (with single unit cell parameter) were selectively incorporated into the NdBaCo2O5 + δ structure while retaining layer-by-layer growth. This work opens routes for systematically depositing heterostructures of these complex oxides with cation order. The electrochemical properties of these epitaxial double perovskite layers are still to be evaluated.

Fig. 13
figure 13

a High-resolution TEM image for a cross section of a 34-nm ordered NdBaCo2O5 + δ film grown by conventional mode. The film/substrate interface is highlighted by two arrows. b Fourier transform where arrows show the doubling of the lattice parameter along the [001] direction. Reproduced with permission from [151] American Chemical Society © 2010

Finally, the first GdBaCo2O5 + δ (GBCO) growth studies by PLD [152] have shown the microstructural complexity of this compound in an epitaxial thin-film form. The epitaxial GBCO films prepared mainly consisted of single- and double-perovskite regions presenting changes in the film orientation from a(b)- to c-oriented upon deposition temperature increase. In addition, the presence of a high density of planar defects, consisting of additional rock-salt GdO layers (as in Ruddlesden Popper phases), was generated by a deviation in the composition of the films. The film electronic conductivities seem to be mainly correlated with the cation composition; so, the larger the deviation from stoichiometric composition, the lower is the conductivity. Despite the presence of defects, the conductivity of the films, with values as high as 800 S/cm at 330 °C in 1 atm O2, is considered very promising for their application as cathodes in intermediate-temperature SOFCs. Although it is clear that further investigations are required, the possibility of tailoring the electronic conductivity by inducing microstructural defects could open a new field of research in thin films.

Another anisotropic material with layered structure which had been proposed as MIEC for IT-SOFC cathode is Sr4Fe6O13, whose structure can be considered as a stacking of SrFeO3 perovskite blocks with FeO6 in octahedral coordination separated by a double layer of FeO4–FeO5 polyhedra. A comprehensive study carried out by Solis et al. [18, 153] showed a strong variation of the transport properties by varying the film thickness of b-axis-oriented epitaxial thin films grown on three different substrates (SrTiO3, NdGaO3 and LaAlO3), showing an unexpected enhancement for strained films. The strain accommodation was found to vary as a function of film thickness and of the substrate material, causing different types of defects in the film microstructure, as well as variations in the oxygen anion content and ordering. Unfortunately, so far, there has been no experimental evidence of enhanced ionic conductivity predicted by theoretical studies [154] for this structure.

Thin-film anodes

In the SOFC anode, the fuel oxidises electrochemically with the oxide ions supplied through the electrolyte membrane. The conditions that the anode material has to fulfil, some of them matching those of the cathode, are mainly that (1) the material should be electronic conducting under the typical reducing conditions for the fuel atmosphere to collect the electrons produced in the fuel oxidation and conduct them to the external circuit; (2) ideally, the material would also be ionic conducting in order to extend the interface between the electrolyte, the electronic conductor and the fuel gas phase (triple-phase boundary); (3) it has to be catalytically active for that oxidation and ideally for hydrocarbon cracking; (4) it should be chemically stable at the extreme conditions of the cathode (to prevent performance loss due to catalyst poisoning by sulphur or graphitic carbon formation).

Current-state SOFCs do not use thin-film anodes. There are very few works dealing with thin-film anodes, and most of them are model studies for the catalytic oxidation of hydrocarbons and the combination of metal oxides (such as doped CeO2) with small charges of metal catalysts (see for instance [155157]). In most of the cases (particularly when using H2 as fuel), it is generally accepted that the thickness influence in the charge transport processes at the anode does not limit the electrochemical performance of the total cell. Therefore, the typical choice for present SOFCs are based in anode-supported cells with the anode being thick enough to mechanically maintain the integrity of the cell. Conventional Ni–YSZ anodes have been prepared by sintering at high temperatures over 1,300 °C. Obviously, this process is not compatible with the preparation of the anode layer on the thin-film electrolyte.

However, recent interest in developing either metal-supported (porous ferritic steel or Ni alloys) [158] or self-supported SOFC platforms, with extremely low thermal mass [1, 159163], requires the use of thin layers for all components. In this sense, there have been some efforts to develop thin-film nanocomposite anodes [156, 164169]. Particularly interesting is the fabrication of Ni:Gd-substituted CeO2 thin-film cermet reported by Infortuna et al. [170], obtained by pulsed laser deposition from the direct ablation of mixed NiO–Gd:CeO2 targets. The cermet showed a stable microstructure with an intimate dispersion of interconnected Ni particles in a porous electrolyte matrix, avoiding Ni coarsening up to an operation temperature of 500 °C. Other attempts to grow thin films suitable as anode material are mainly related to Pt [171] or Ni deposition [172]. In some cases and particularly when using metallic supports, it is necessary to use thin diffusion barriers to avoid Cr or Fe diffusion into the Ni-containing anode layer [173] or additional thin activation layers, with a finer porous structure which guarantee the catalytic activity for the hydrocarbon oxidation [174]. Concerning the anode element, there is still substantial place for the development of new Ni-free compositions and structures that provide high power density at low temperature with low degradation rate and tolerance to sulphur poisoning and coke formation. In this sense, perovskite-related single-phase materials such as Y-substituted SrTiO3 [175179], La0.75Sr0.25Cr0.5Mn0.5O3 [180], tungstate bronzes or pyrochlores [181] have demonstrated considerable electronic conductivity and redox capabilities while maintaining sulphur tolerance and are excellent candidates to be explored in thin-film form.

Conclusions and future perspectives

At the moment, research on epitaxial thin-film materials for electrolyte and electrode applications truly constitutes a ‘hot topic’ within the solid-state materials community, clearly reflected by the increasing number of articles related to the area which have been published in the last years. Most of the studies included in the present review concern the use of epitaxial films for fundamental research on the intrinsic properties of materials for IT-SOFCs. In addition, it has been shown that important materials properties, such as the oxide ionic and electronic transport, along with the oxygen surface exchange, can be tailored by selecting the substrate and the deposition conditions used for preparing epitaxial films and heterostructures.

Concerning SOFC devices, the present application of thin films is limited to polycrystalline or highly textured thin deposits. However, there are considerable expectations of incorporating epitaxial films and heterostructures in future monolithic SOFC devices and particularly in μSOFC platforms. In μSOFCs, the reduction of the thickness of the overall cell components allows, on the one hand, minimisation of the ohmic losses due to the limited charge transport through the elements, particularly at lower operating temperatures (from 400 to 600 °C) and, on the other hand, reduction of the thermal mass of the active cell volume, reducing as a result the thermal energy losses and gaining portability. Several comprehensive reviews dealing with μSOFCs’ fundamental and more technological aspects can be found in [1, 159162, 182184].

In summary, although the number of research studies concerning epitaxial thin films for SOFC materials is still rather limited, especially in the case of actual SOFC devices, it can be already foreseen that the versatility of the epitaxial growth for both fundamental studies and final applications is enormous, and it is expected to be an area of intense research efforts during the next years.